Choi 2018

Subscriber access provided by University of Winnipeg Library
A: Spectroscopy, Molecular Structure, and Quantum Chemistry

Tuning Hydrogenated Si, Ge, and SiGe Nanocluster Properties
Using Theoretical Calculations and a Machine Learning Approach
Yeseul Choi, and Andrew J Adamczyk
J. Phys. Chem. A, Just Accepted Manuscript • DOI: 10.1021/acs.jpca.8b09797 • Publication Date (Web): 28 Nov 2018
Downloaded from http://pubs.acs.org on November 30, 2018
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,

Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.
Page 1 of 48 The Journal of Physical Chemistry
1
2
3
4 1 Tuning Hydrogenated Si, Ge, and SiGe Nanocluster Properties
5
6 2 using Theoretical Calculations and a Machine Learning
7
8 3 Approach
9
10 4 Yeseul Choi1, Andrew J. Adamczyk1,*
11 5
12 6 1Auburn University, Department of Chemical Engineering, Auburn, AL 36849, USA.
13 7
14 8 *Corresponding author e-mail: aja0056@auburn.edu.
15
16
17 9 Abstract
18 10
19 11 There are limited studies available that predict the properties of hydrogenated silicon-germanium
20
21 12 (SiGe) clusters. For this purpose, we conducted a computational study of 46 hydrogenated SiGe
22
23
13 clusters (SixGeyHz, 1<X+Y≤6) to predict the structural, thermochemical, and electronic properties.
24 14 The optimized geometries of the SixGeyHz clusters were investigated using quantum chemical
25
26 15 calculations and statistical thermodynamics. The clusters contained 6 to 9 fused Si-Si, Ge-Ge, or
27 16 Si-Ge bonds, i.e., bonds participating in more than one 3- to 4-membered rings, and different
28
29 17 degrees of hydrogenation, i.e., the ratio of hydrogen to Si/Ge atoms varied depending on cluster
30
18 size and degree of multifunctionality. Our studies have established trends in standard enthalpy of
31
32 19 formation, standard entropy, and constant pressure heat capacity as a function of cluster
33
34 20 composition and structure. A novel bond additivity correction model for SiGe chemistry was
35 21 regressed from experimental data on 7 acyclic Si/Ge/SiGe species to improve the accuracy of the
36
37 22 standard enthalpy of formation predictions. Electronic properties were investigated by analysis of
38 23 the HOMO–LUMO energy gap to study the effect of elemental composition on the electronic
39
40 24 stability of SixGeyHz clusters. These properties will be discussed in the context of tailored
41
25 nanomaterials design and generalized using a machine learning approach.
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 1|Page
57
58
59
60 ACS Paragon Plus Environment
The Journal of Physical Chemistry Page 2 of 48
1
2
3
4 26 Introduction
5
6 27 Studies of semiconducting silicon-germanium (SiGe) materials are of technological interest
7 28 because of their practical application in the microelectronics industry.1 Moreover, SiGe clusters
8
9 29 have attracted great interest for their use in optoelectronic, sensor, and photovoltaic applications.2-
10 30 6 Understanding semiconducting nanomaterials formation from the pyrolysis of mixtures of silane
11
12 31 (SiH4) and germane (GeH4) at even the mildest conditions is still incomplete.7-12 Homogenous
13
32 gas-phase nanomaterials formation is a complex phenomenon in which hundreds, or possibly
14
15 33 thousands of species, undergo simultaneous reaction. During the chemical vapor deposition of
16
17 34 SiGe semiconducting nanomaterials, surface reactions play an important role. However,
18 35 undesired defects can arise in semiconductor processing because these SiGe clusters deposit
19
20 36 on the growing substrate. Since these clusters are important for the fine processing of
21 37 semiconductors and the synthesis of novel materials, computational modeling can play a very
22
23 38 important role in narrowing the gap between controlled experimental studies and practical
24
25
39 operating conditions. Similarly, intentional synthesis of SiGe semiconducting nanomaterials in the
26 40 gas phase can benefit from an improved mechanistic understanding of formation to tailor efforts
27
28 41 in materials design, particularly for self-assembling molecular systems and nanocomposites.
29 42
30
31 43 Due to tunable semiconductor properties, Si, Ge, and SiGe clusters have also attracted great
32
44 interest for the development of new materials in nanoscale applications as fundamental building
33
34 45 blocks.13 Both Si and Ge clusters have widely been studied because the structure and bonding of
35
36 46 bulk Ge materials are very similar to that of bulk Si materials.14 Pristine Si and Ge clusters without
37 47 hydrogen content are chemically reactive and thus not suitable as a building block for
38
39 48 self-assembled materials. However, this reactivity can be reduced with surface passivation by
40 49 hydrogen or other suitable functional groups, such as alkyl functionalities. Si clusters have been
41
42 50 studied extensively for their promising structural, thermochemical, and electronic properties.15-21
43
44
51 For Ge clusters, there have been reported cage types of pure Ge structures with metal-doping.22-24
45 52 Most Ge cluster studies have been conducted to investigate the geometric strain effect of clusters
46
47 53 upon increased Ge content for medium to large cluster sizes.6 However, to the best of our
48 54 knowledge, analogous studies have not been reported for small- to medium-sized hydrogenated
49
50 55 SiGe clusters.
51
56
52
53 57 Furthermore, nanocrystals of Ge have received significant interest in recent years.5
54
55 58 Self-organized quantum dots of Ge were grown on Si substrates.25 It was observed in Si/Ge
56 2|Page
57
58
59
1
2
3 59 superlattices26, 27 and in Ge quantum dots grown on Si25 that interdiffusion between Si and Ge
4
5 60 may occur to form alloys under certain growth conditions. Detailed knowledge of the
6
61 thermodynamics and the nature of Ge–Si bonding is still needed to understand the spontaneous
7
8 62 processes leading to the formation of self-organized structures. Compared to the vast data
9
10 63 available on solid-state materials, theoretical solid-state studies on materials possessing Si-Ge
11 64 bonds and comprehension of the SiGe chemistry, especially for small clusters, are very rare.28-31
12
13 65 The limited results which are available on such model clusters are confined to thermodynamic
14 66 investigations of the clusters of small sizes less than four atoms,32, 33 measurement of optical
15
16 67 properties of SiGe materials,34 and a few advanced ab initio calculations on SiGe dimers.35, 36
17
68 Structural characteristics were also determined theoretically for several selected larger clusters
18
19 69 by applying semi-empirical methods which include tight-binding molecular dynamics
20
21 70 approaches,37, 38 density functional theory (DFT),39 and Møller–Plesset second order perturbation
22 71 (MP2) theories.40 So far, no theoretical thermodynamic data, which could indirectly validate the
23
24 72 calculated structures through comparison against the existing experiments, are available. Thus,
25
73 detailed theoretical studies connecting the structures, bonding, and thermodynamic properties of
26
27 74 Si, Ge, and SiGe clusters are critically needed.
28
29 75
30 76 Recently, automated network generation techniques41 have allowed the kinetics of inorganic
31
32 77 cluster and nanoparticle formation, such as Si clusters and nanoparticles, to be described at the
33 78 mechanistic level.42, 43 Rate coefficients must be estimated for every elementary step comprising
34
35 79 the mechanistic model, and kinetic correlations are used to make this tractable. One common
36
80 method for predicting activation barriers (Ea) is the Evans-Polanyi correlation; however, these
37
38 81 structure-activity correlations require detailed thermochemical information for each reacting
39
40 82 species. Recently, the existing group additivity database8 for the prediction of thermochemical
41 83 properties of hydrogenated silicon clusters was revised and augmented with new atom-centered
42
43 84 groups, ring corrections, and bond-centered groups to accurately capture more complex species.
44 85 Conversely, there are limited studies available that predict the thermochemical properties of SiGe
45
46 86 and Ge clusters, which is the next step for expanding our thermochemistry database for
47
48
87 semiconducting nanoparticle formation.
49
50 88 For this purpose, we conducted a computational study of hydrogenated Si, Ge, and SiGe alloy
51
52 89 clusters (SixGeyHz, 1<X+Y≤6) to predict structures, thermochemistry, and electronic properties.
53
90 This paper presents the thermochemical properties of 46 cyclic and polycyclic Si, Ge, and SiGe
54
55
56 3|Page
57
58
59
1
2
3 91 clusters and 7 acyclic Si, Ge, and SiGe species, i.e., standard enthalpies of formation, standard
4
5 92 entropy values, and constant pressure heat capacities, and specifically examines both
6
93 multifunctional and monofunctional molecules containing between one and six Si and/or Ge
7
8 94 atoms. The hydrogenated clusters in this study involved different degrees of hydrogenation, i.e.,
9
10 95 the ratio of hydrogen to Si and Ge atoms varied widely depending on the size of the cluster and/or
11 96 degree of multifunctionality. Species containing different numbers of fused rings comprised of
12
13 97 three to four Si or Ge atoms were considered. The composite method of G3//B3LYP44, 45 was
14 98 used to calculate the electronic energy, and then statistical thermodynamics was applied to all
15
16 99 the hydrogenated Si, Ge, and SiGe clusters to incorporate temperature effects. Enthalpies of
17
100 formation at 1 atm and 298 K were calculated using atomization energies and corrected with a
18
19 101 novel bond additivity correction model. Standard entropies and constant pressure heat capacities
20
21 102 were calculated using a temperature-dependent scaling factor for the vibrational frequencies to
22 103 account for anharmonicity. Our studies have established trends in thermodynamic properties
23
24 104 (standard enthalpy of formation (ΔHof), standard entropy (So), and constant pressure heat capacity
25
105 (Cp)), as a function of cluster composition and structure. Furthermore, we compared
26
27 106 HOMO-LUMO energy gaps and HOMO and LUMO electron distributions in order to gain insight
28
29 107 into the electronic stability of the hydrogenated Si, Ge, and SiGe clusters. Quantum chemical
30 108 parameters such as electronic chemical potential , global hardness , and the softness  were
31
32 109 also calculated to provide valuable information about chemical stability. These quantum chemical
33
110 parameters were generalized using a machine learning approach to assess charge transfer during
34
35 111 molecular interaction of hydrogenated Si, Ge, and SiGe clusters in the gas phase.
36
37
38 112 COMPUTATIONAL METHODOLOGY
39
40 113 Quantum chemical calculations were performed with the Gaussian 16 software.46 All electronic
41
42 114 energies for the hydrogenated Si, Ge, and SiGe clusters and acyclic species were calculated
43 115 using the G3//B3LYP composite method,35, 44 which uses B3LYP/6-31g(d) geometries and
44
45 116 higher-level corrections based on single point energies. To assess different levels of theory, we
46 117 employed the Gaussian 16 software to perform quantum chemical calculations using the
47
48 118 CBS-QB3, G3//B3LYP and G4//B3LYP composite methods. The primary difference between the
49
50
119 Gn and CBS methods is how the correlation energy is estimated. The Gn methods assume basis
51 120 set additivity and add an empirical correction to recover part of the remaining correlation energy.
52
53 121 The complete basis set (CBS) procedures, on the other hand, attempt to perform an explicit
54 122 extrapolation of the calculated values.47 All electronic energies for the hydrogenated Si, Ge, and
55
56 4|Page
57
58
59
1
2
3 123 SiGe acyclic species in this study were calculated using these three levels of theory.
4
5 124
6
125 The optimized structures for all 46 hydrogenated Si, Ge, and SiGe clusters investigated in this
7
8 126 study are depicted in Figure 1. The hydrogenated clusters of this study can exist in the singlet
9
10 127 state and triplet state.8, 48-50 As shown in Table 1, using the G3//B3LYP method, triplet-singlet
11 128 splitting values of linear and cluster species were investigated. These calculated triplet-singlet
12
13 129 splitting values suggest that the singlet potential energy surface is significantly lower in energy
14 130 than the triplet potential energy surface. Thus, for all results reported in this study, the electronic
15
16 131 wave functions for the hydrogenated Si, Ge, and SiGe clusters were optimized in the singlet state.
17
132 Geometries and harmonic vibrational frequencies are confirmed local minima on the singlet
18
19 133 potential energy surface, i.e., all of the vibrational frequencies are real. It is well-established that
20
21 134 hydrogenated Si, Ge, and SiGe nanostructures pass through metastable configurations (or
22 135 transient chemical species) before reaching a global minimum from molecular dynamics
23
24 136 simulations,51 but detailed knowledge of the structure and thermochemistry of a wide range of
25
137 hydrogenated clusters is still needed. The harmonic vibrational frequencies and zero-point
26
27 138 vibrational energy (ZPE) were linearly scaled by a temperature-dependent scaling factor of 0.98,
28
29 139 respectively, to account for anharmonicity in the normal vibrational modes as a function of
30 140 temperature as suggested by Scott and Radom and Alecu and co-workers.52, 53 Using
31
32 141 conventional statistical thermodynamics, molecular partition functions based on the harmonic
33 142 oscillator and rigid rotor approximations were used to calculate thermodynamic properties as a
34
35 143 function of temperature.
36
144
37
38 145 A closer investigation of the 46 cluster structures in this study reveals no dangling Si-Si, Ge-Ge,
39
40 146 and Si-Ge σ bonds capable of internal rotation for the clusters with the exception of the substituted
41 147 trigonal planar geometry. Aside from the temperature-dependent scaling factor, anharmonic
42
43 148 movements in torsional vibrational modes for the linear chemical species (Si2H6, Ge2H6, SiGeH6,
44 149 Si3H8, and Ge3H8) and the substituted trigonal planar cluster geometries were not treated.
45
46 150 Similarly, anharmonic small ring movements (e.g., the pseudorotation of cyclopentasilanes and
47
48
151 the ring puckering of cyclotetrasilanes incorporated into the multifunctional polycyclic structures)
49 152 were not treated aside from the temperature-dependent scaling factor.54, 55 The protocol in our
50
51 153 study was implemented because (1) there are reduced anharmonic small ring movements for the
52 154 more rigid structures in this study, which was verified by the animation of key vibrational modes,
53
54 155 (2) the calculation of a revised partition function to account for anharmonic torsional modes and
55
56 5|Page
57
58
59
1
2
3 156 small ring movements was beyond the scope of this study. Enthalpy, H, and entropy, S, are
4
5 157 calculated using standard formulas.56 Calculation of thermochemical properties was performed
6
158 automatically using the CalcTherm script, which interfaces with electronic structure codes to
7
8 159 provide thermochemical properties (S, Cp , H) of individual species at elevated temperatures.57
9
10 160 The external symmetry numbers for the hydrogenated Si-Ge clusters examined in this study
11 161 impact the molecular partition function for rotation and reduce the rotational entropy by an amount
12
13 162 equivalent to R ln σrot,56 where σrot is the external symmetry number of the molecule and R is the
14 163 ideal gas constant.
15
16 164
17
165 The enthalpy of formation of a given molecule SixGeyHz can be calculated from its atomization
18
19 166 energies using Eq (1):8
20
21 167
22 ∆𝐻°𝑓,298(𝑆𝑖𝑥𝐺𝑒𝑦𝐻𝑧) = [𝑥∆𝐻°𝑓,298 (𝑆𝑖) + 𝑦∆𝐻°𝑓,298 (𝐺𝑒) + 𝑧∆𝐻°𝑓,298 (𝐻)] ― ∆𝐻°𝑎,298 (𝑆𝑖𝑥𝐺𝑒𝑦𝐻𝑧) (1)
23 168
24 169
25
26
170 where the formation enthalpies of atomic silicon, germanium and hydrogen are the experimental
27 171 values obtained from the JANAF tables (∆H◦f,298(Si) = 450 kJ mol-1, ∆H◦f,298(Ge) = 372 kJ mol-1,
28
29 172 ∆H◦f,298(H) = 217.999 kJ mol-1) and ∆H◦f,298(SixGeyHz). The atomization energy defined as the
30 173 enthalpy change upon decomposition of a molecule into its component atoms can be evaluated
31
32 174 using Eq (2):
33
175
34
35 176 ∆𝐻°𝑎,298(𝑆𝑖𝑥𝐺𝑒𝑦𝐻𝑧) = [𝑥𝐻𝑎,298 (𝑆𝑖) + 𝑦𝐻𝑎,298(𝐺𝑒) + 𝑧𝐻𝑎,298(𝐻)] ―𝐻298(𝑆𝑖𝑥𝐺𝑒𝑦𝐻𝑧) (2)
36
37 177
38 178 where H298(Si), H298(Ge) and H298(H) are the enthalpies of atomic silicon, germanium and
39
40 179 hydrogen at 298 K, respectively, and H298(SixGeyHz) is the enthalpy of SixGeyHz at the same
41 180 temperature. These enthalpies can be calculated as the sum of the electronic energies (Eel), zero
42
43 181 point energies (ZPE), and thermal corrections (Evib298, Etrans298, and Erot298) at 298 K, as follows
44
182 from canonical molecular partition functions assuming an ideal gas at 1 atm using Eq. 3:
45
46 183
47
48 184 H 298  E el  ZPE  E 298
vib  E trans  E rot  PV
298 298
(3)
49
50 185
51
52 186 All of the quantities on the right-hand side of Eq 3 are obtained from quantum chemical
53 187 calculations, and the standard enthalpy of formation of SixGeyHz is then calculated.
54
55 188
56 6|Page
57
58
59
1
2
3 189 The isodesmic bond additivity correction (BAC) proposed by Petersson et al.58 and applied to
4
5 190 silicon hydride chemistry by Wong et al. was extended in our calculations of standard enthalpy of
6
191 formation for SiGe and Ge species. This approach uses a set of reference molecules that have
7
8 192 experimental data available which then are compared to G3//B3LYP enthalpies of formation from
9
10 193 homodesmotic reactions to calculate a set of correction parameters. The inclusion of these
11 194 correction parameters was shown to lead to values that very closely approximate standard
12
13 195 enthalpies of formation based on available experimental data and data calculated using the
14 196 method of homodesmotic reactions.59 The novel BAC parameters used in this study to calculate
15
16 197 enthalpies of formation for Si, Ge, and SiGe species are regressed and presented in the Results
17
198 and Discussion section and follow Eq 4.
18
19 199
20
21 200 ΔHf,298 (BAC)  ΔHf,298 (calculated)   N i BACi (4)
22 i
23
24 201 BACi is the BAC parameter of a certain bond type i, and the standard enthalpy of formation
25
202 estimated from BACs, ΔHof,298 (BAC), can be defined as the standard enthalpy of formation
26
27 203 calculated on the basis of atomization energies, ΔHof,298 (calculated), corrected by the
28
29 204 summation of the BACi parameters multiplied by the number of bonds of that type (Ni).
30 205
31
32 206 In order to test the accuracy of our calculations, calculations were carried out on small acyclic Si,
33 207 Ge, and SiGe hydrides using the CBS-QB3, G3//B3LYP, and G4//B3LYP methods and
34
35 208 summarized in Table 2. The calculated results from the G3//B3LYP and G4//B3LYP composite
36
209 methods were found to be in reasonable agreement with available experimental data for standard
37
38 210 enthalpy of formation at 298 K. The performance summary for prediction of thermochemical
39
40 211 properties for small acyclic Si, Ge, and SiGe hydride chemistries indicated that the G3//B3LYP
41 212 and G4//B3LYP composite methods outperform the CBS-QB3 method on estimating standard
42
43 213 enthalpy of formation when compared to available experimental data. Calculation of the standard
44 214 enthalpy of formation at 298 K has been underestimated with CBS-QB3, G3//B3LYP and
45
46 215 G4//B3LYP composite methods. In the case of estimating standard entropy values of the acyclic
47
48
216 species, all predicted standard entropies were estimated between 0.5 and 3.8 J mol-1 K-1 in
49 217 average absolute deviation when compared to available experimental data. For constant pressure
50
51 218 heat capacity, however, the calculations for all three methods were very accurate with a highest
52 219 average absolute deviation of 1.1 J mol-1 K-1 when compared to available experimental data.
53
54 220
55
56 7|Page
57
58
59
1
2
3 221 Analogously, a previous study on silicon hydrides from our group8 compared W-1 and G3//B3LYP
4
5 222 standard enthalpies of formation at 298 K to available experimental data. The G3//B3LYP
6
223 composite method agrees with available experimental data within an average absolute deviation
7
8 224 of 1.0 kcal mol-1 while the W-1 method captures available experimental data within an average
9
10 225 absolute deviation of 2.0 kcal mol-1. The W-1 method was developed to be an affordable and
11 226 accurate method for the determination of thermochemistry; however, the improved predictions by
12
13 227 the G3//B3LYP method can be attributed to the use of a higher level correction based upon a
14 228 regression of correction parameters from an experimental data set of 299 energies containing
15
16 229 enthalpies of formation, ionization potentials, electron affinities, and proton affinities.44, 60
17
230 Additionally, the G3//B3LYP method was a reasonable choice because the cluster property data
18
19 231 set from this study is intended to be used in conjunction with an existing G3//B3LYP database
20
21 232 developed by our group7-12 for the estimation of silicon hydride thermochemical properties. The
22 233 G4//B3LYP method did exhibit superior accuracy for the small acyclic species examined with
23
24 234 available experimental data; however, we experienced significant self-consistent field (SCF)
25
235 energy convergence issues for structures larger than four Si and/or Ge atoms in the latest revision
26
27 236 of the Gaussian software and this method was not pursued.
28
29 237
30 238 Any chemical system (e.g., an atom, molecule, ion, or radical) is characterized by its electronic
31
32 239 chemical potential, , and by its absolute hardness, . Thus, the calculated quantum chemical
33
240 parameters such as the highest occupied molecular orbital energy EHOMO, the lowest unoccupied
34
35 241 molecular orbital energy ELUMO, energy gap E, electronic chemical potential , global hardness
36
37 242 , and the softness  were calculated in our study. The concept behind the derivation of these
38 243 parameters is related to each other through frontier molecular orbital theory,61-64 and this concept
39
40 244 can be approximated as Eqs 5 and 6.
41
42
245
43 𝐼+𝐴
246 ―𝜇 = =𝜒 (5)
44 2
45 247
46
47 248
48 𝐼―𝐴
49 249 𝜂= 2 (6)
50
51 250
52 251 Here I is the ionization potential and A is the electron affinity. The inverse values of the global
53
54 252 hardness are designated as the softness 𝜎 = 1 𝜂. According to Koopmans’ theorem, the frontier
55
56 8|Page
57
58
59
1
2
3 253 orbital energies are given by -EHOMO = I and -ELUMO= A. It is well-known and controversial that
4
5 254 application of Koopmans’ theorem to Kohn–Sham (KS) Density Functional Theory (KS-DFT)
6
255 requires a tuning procedure to be able to "impose" Koopmans' theorem on DFT approximations,
7
8 256 thereby improving many of its related predictions in actual applications.65, 66 Here hybrid
9
10 257 functionals systematically calculate HOMO energies that underestimate the first ionization
11 258 potential values by several electron volts. Nevertheless, these tabulated quantum chemical
12
13 259 parameters can be used in two possible ways: as a rank ordering of similar acids (electrophiles)
14 260 or bases (nucleophiles) to predict relative properties or as a source of values to use in relevant
15
16 261 equations such as Eq 7. If two systems or molecules, A and B are brought together, electrons will
17
18 262 flow from that of lower  to that of higher , until the chemical potentials become equal. As a first
19 263 approximation, the (fractional) number of electrons transferred, N, will be given by Eq 7. The
20
21 264 difference in electronegativity drives the electron transfer, and the sum of the hardness
22
265 parameters acts as a resistance. This reactivity index was then generalized beyond the species
23
24 266 in this study using a machine learning approach based on multiple linear regression and detailed
25
26 267 sensitivity analysis.
27 268
28
𝜒𝐵 ― 𝜒𝐴
29 269 ∆𝑁 = 2(𝜂𝐵 + 𝜂𝐴) (7)
30
31 270
32
33 271 Results and Discussion
34
35 272
36
37 273 Structures and Vibrational Frequencies
38
39 274 The structures for the 46 hydrogenated Si, Ge, and SiGe clusters and 7 acyclic Si, Ge, and SiGe
40 275 species that were investigated in the present study were optimized using the B3LYP/6-31G(d)
41
42 276 level of theory. The optimized structures for all of the clusters showed complex polycyclic or cyclic
43 277 nature and a varying level of surface passivation with hydrogen atoms, as illustrated in Figure 1.
44
45 278 High-energy sterically strained structural isomers were calculated in this study to capture the
46
279 diverse range of strain energies possible in hydrogenated Si, Ge, and SiGe clusters. Acyclic (or
47
48 280 linear) Si, Ge, and SiGe structures were calculated for species comprised of one to three Si or
49
50 281 Ge atoms. For the cluster structures, hydrogenated trigonal planar, trigonal pyramidal, substituted
51 282 trigonal planar, trigonal bipyramidal, and prismane geometries comprised of varying numbers of
52
53 283 three- and four-membered rings were calculated. All electronic wavefunctions for the structures
54
284 were optimized in a singlet state. We also calculated all structures in this study in the triplet states
55
56 9|Page
57
58
59
1
2
3 285 (see Table 1 for selected structures). It was observed that the clusters and acyclic species
4
5 286 changed geometry significantly upon excitation to the triplet state. In the case of trigonal planar
6
287 structures going to the triplet state from the singlet state, the structures optimized to a linear
7
8 288 geometry which has a higher standard enthalpy of formation than its analogue in the singlet state.
9
10 289 For the three dimensional structures such as trigonal pyramidal, trigonal bipyramidal, and
11 290 prismane, the structures transformed to less stereoscopic shapes which appeared similar to
12
13 291 cyclohexane in geometry. This strong change in geometry indicated that the structures in this
14 292 study are very stable in singlet ground state, and thus require significant structural rearrangement
15
16 293 to find a stable minima on the triplet potential energy surface. Electronic excitation to a higher
17
294 spin state may be a viable means to create more reactive intermediates due to the conformational
18
19 295 changes in geometry on the path to the triplet state from the singlet state.
20
21 296
22 297 Comparison of our predicted values to experiment for geometry parameters are presented here
23
24 298 for the acyclic species. For silane, the B3LYP/6-31G(d) predicted Si-H bond distance is 1.486 Å
25
26
299 and the experimental value is 1.480 Å.67 For germane, the predicted Ge-H bond distance is 1.542
27 300 Å and the experimental value is 1.525 Å.68 For disilane, the predicted Si-Si and Si-H bond lengths
28
29 301 were 2.350 Å and 1.489 Å, respectively, and experimental values were 2.331 Å and 1.492 Å,
30
31
302 respectively. The HSiSi and HSiH bond angles were predicted as 110.6 and 108.3 degrees,
32 303 respectively, and experimental values were 110.3 and 108.6 degrees, respectively. For
33
34 304 digermane, the predicted Ge-Ge and Ge-H bond lengths were 2.448 Å and 1.546 Å, respectively,
35
305 and experimental values were 2.403 Å and 1.541 Å, respectively. The HGeGe and HGeH bond
36
37 306 angles were predicted as 110.7 and 108.2 degrees, respectively, and experimental values were
38
39 307 112.3 and 106.4 degrees, respectively. For H3SiGeH3, the predicted Si-Ge, Si-H, and Ge-H bond
40 308 lengths were 2.398 Å, 1.488 Å, and 1.546 Å, respectively, and the experimental values were 2.358
41
42 309 Å, 1.494 Å, and 1.538 Å, respectively. The HSiH and HGeH bond angles were predicted as 108.5
43
44 310 and 107.9 degrees, respectively, and experimental values were 108.8 and 108.3 degrees,
45 311 respectively.69 All of our predicted values for geometry parameters are very well matched to the
46
47 312 experimental data for the acyclic Si, Ge, and SiGe species in this study.
48 313
49
50 314 Using the nomenclature introduced in Figure 1, an analysis of the geometry parameters for the
51
315 cluster species are presented herein. For trigonal planar Si3H6 (T-0), the equilateral triangle
52
53 316 structure is suggested with a Si-Si bond length of 2.345 Å and a Si-H bond length of 1.486 Å.
54
55 317 Substituting Ge atoms systematically for Si atoms in the T-0 geometry, all the mixed SiGe and a
56 10 | P a g e
57
58
59
1
2
3 318 pure Ge trigonal planar geometries were created. Addition of a Ge atom to the T-0 geometry
4
5 319 increases the length of all the bonds mildly, thereby increasing the size of the full cluster where
6
320 the fully substituted Ge cluster is the largest in geometric dimensions. For the T-1 geometry, the
7
8 321 Ge-Si bond length is 2.405 Å, Si-Si bond length is 2.348 Å, Ge-H bond length is 1.541 Å, and
9
10 322 Si-Ge-Si apex angle is 58.6 degrees. As observed with the trigonal planar structural series, all
11 323 other structural geometries (i.e., substituted trigonal planar, trigonal pyramidal, trigonal
12
13 324 bipyramidal, and prismane) showed a similar trend of expanding bond lengths when exchanging
14 325 a Si atom with a Ge atom. Although not presented in Figure 1, it is noteworthy to discuss the S-0
15
16 326 geometry, or nearly planar cyclic rhombus structure of four Si atoms, which has a Si-Si bond
17
327 length of 2.371 Å and a Si-H bond length of 1.492 Å. With the angles of the Si-Si-Si bonds at 87.7
18
19 328 and 92.3 degrees and all Si atoms possessing an sp3 hybridized center, this cyclic structure shows
20
21 329 a slightly puckered character and the four-membered ring is not completely planar. Interestingly,
22 330 the square planar Si4H8 is the only structure in the geometry series which showed a stable
23
24 331 minimum on the singlet potential energy surface with all real vibrational frequencies. Structures
25
332 comprised of one four-membered ring and any level of Ge content were unstable, i.e., all
26
27 333 structures were found to be higher-order saddle points on the potential energy hypersurface with
28
29 334 imaginary vibrational frequencies. A conformational search revealed that a substituted trigonal
30 335 planar structure in which a hydrogen in the trigonal plane structure is substituted with a silyl or
31
32 336 germyl group was a more stable minimum on the singlet potential energy surface with all real
33 337 vibrational frequencies. We found that the substituted trigonal planar (ST-0) geometries are more
34
35 338 stable than the square planar (S-0) geometries, where a smaller three-membered ring in the ST-
36
339 0 geometry is preferred over a larger four-membered ring in the S-0 geometry. All ST-0
37
38 340 geometries with varying levels of Si and Ge content have real vibrational frequencies. In the case
39
40 341 of Si4H8 (ST-0), the average of Si-Si and Si-H bond lengths are 2.346 Å and 1.486 Å, respectively,
41 342 and the angle between the trigonal plane and the Si atom of the silyl group is 120.6 degrees.
42
43 343 These bond distances and angles in the ST-0 structural series are slightly increased as the Si
44 344 atoms are replaced with Ge atoms.
45
46 345
47
48
346 Compared to the Si atom, the Ge atom has a full 3d shell of 10 electrons and significantly more
49 347 electrons than the Si atom; however, the respective bond lengths upon Ge substitution are only
50
51 348 increased by +3.7% going from the Si-H to Ge-H substitution, respectively, and +2.4% going from
52 349 Si-Si to Si-Ge substitution, respectively. Another reason for the observed higher stability for the
53
54 350 ST-0 geometry than the S-0 geometry can be seen by the fact that the ST-0 geometry has fewer
55
56 11 | P a g e
57
58
59
1
2
3 351 overall spatial constraints than the S-0 geometry. In other words, the ST-0 geometry still has a
4
5 352 torsional degree of freedom in vibrational modes for the substituted silyl or germyl group which is
6
353 lost upon a four-membered ring formation. This explanation could also attribute why the ST-0
7
8 354 geometry is more stable than the TP-0 geometry, which also lacks any torsional degrees of
9
10 355 freedom in vibrational modes. The trigonal pyramidal geometry (TP-0) has on average a shorter
11 356 Si-Si bond length at 2.327 Å and Si-H bond length at 1.479 Å compared to Si clusters in the other
12
13 357 geometry series. In the TP-1 structure, the Si-Si bond length on average is 2.332 Å and the Si-Ge
14 358 bond length is 2.399 Å. The fact that the TP-0 geometry has more contracted Si-Si bond distances
15
16 359 than the ST-0 geometry supports the stable nature of ST-0 geometry due to an overall lack of
17
360 polycyclic nature.
18
19 361
20
21 362 Several structures in this study were initially hypothesized to have both pentacoordinated and
22 363 hexacoordinated Si and Ge centers; however, only stable clusters comprised of hexacoordinated
23
24 364 Si and Ge centers were isolated. This type of bonding behavior indicates that both Si and Ge
25
365 centers would exhibit sp3d or sp3d2 hybridization, respectively. In its least strained hypervalent
26
27 366 form, sp3d hybridized Si and Ge centers will form covalent bonds with five neighboring atoms in
28
29 367 a trigonal bipyramidal electron pair coordination. These sp3d hybridized centers were explored for
30 368 the square bipyramidal (SBP) geometry series. The sp3d2 hybridized Si and Ge centers will form
31
32 369 covalent bonds with six neighboring atoms in an octahedral electron pair coordination. These
33 370 sp3d2 hybridized Si and Ge centers were explored for trigonal bipyramidal (TBP) geometries.
34
35 371 Structures comprised of pentacoordinated Si and Ge were found to be the unstable in the square
36
372 bipyramidal geometries and these structures favored prismane geometries (Pri) that were instead
37
38 373 comprised of sp3 hybridized Si and Ge atoms. These SBP structures comprised of
39
40 374 pentacoordinated Si and Ge centers were unstable, i.e., all structures were found to be
41 375 higher-order saddle points on the potential energy hypersurface comprised of imaginary
42
43 376 vibrational frequencies. The unstable SBP structure constructed to have sp3d hybridization is
44 377 supported with similar work done for Si and Ge complexes where Si complexes were found to be
45
46 378 more stable in the hypervalent state with a hexacoordinated complex rather than a
47
48
379 pentacoordinated complex.70 This tendency to form a hexacoordinated Si or Ge centers is further
49 380 supported by our stable trigonal bipyramidal geometries which exhibit a strained hexacoordinated
50
51 381 Si or Ge center in the trigonal center plane of the cluster. For the molecules in our study, results
52 382 show that a similar preference for hypervalent bonding behavior observed for the Si centers also
53
54 383 occurs for the Ge centers. This bonding behavior is likely due to the presence of a complete 3d
55
56 12 | P a g e
57
58
59
1
2
3 384 shell of 10 electrons for the Ge atom which would exhibit more facility to form hybridization
4
5 385 involving the d orbital than the Si atom. In this study, all of the Si and Ge atoms are passivated
6
386 with hydrogen or bonded to other Si or Ge atoms to be in the most stable sp3 and sp3d2
7
8 387 hybridization states. As with the trigonal planar (T-0), substituted trigonal planar (ST-0), and
9
10 388 trigonal pyramidal (TP) geometries, the expansion of bond lengths and bond angles upon
11 389 substitution of a Ge atom for a Si atom was also observed for trigonal bipyramidal (TBP) and
12
13 390 prismane (Pri) geometries. There are studies in the literature for so-called “ultrastable silicon
14 391 nanoclusters”, or hydrogenated pure silicon prismanes comprised of up to 18 silicon atoms. Katin
15
16 392 et al.71 compared the electronic, optical properties, and kinetic stability of Si18H12 with the pristine
17
393 silicon prismane and prismanes embedded with additional C, Si, and Ge atoms. Comparison of
18
19 394 this theoretical study with our current study, clusters with a higher density of atoms are subject to
20
21 395 have larger spatial confinements which resulted in shorter Si-Si, Si-Ge, and Ge-Ge bond lengths.
22 396 A similar tendency in bond contraction can be observed by changing the Si atom to a Ge atom in
23
24 397 the structures, as opposed to embedding an additional atom in the center of the polycyclic
25
398 structure.
26
27 399
28
29 400 At the level of theory considered in our study, modeling results have all real vibrational frequencies
30 401 and represent stable minima on the potential energy surface. Experimental spectroscopic data
31
32 402 for the vibrational frequencies of hydrogenated Si, Ge, and SiGe clusters are limited. The
33 403 unscaled harmonic vibrational frequencies for linear silicon hydrides (SiH4 and Si2H6) and
34
35 404 germanium hydrides (GeH4 and Ge2H8) calculated using two different level of theories (G3//B3LYP
36
405 and G4//B3LYP) were compared against available spectroscopic experimental data8, 68 in Table
37
38 406 3. The unscaled harmonic frequencies for SiH4, Si2H6, GeH4, and Ge2H6 were determined to have
39
40 407 mean percentage deviations from experimental values of -1.7, -1.8, +0.7, and +5.8 %, respectively,
41 408 at the G3//B3LYP level of theory. The G4//B3LYP method predicts slightly more accurate
42
43 409 vibrational frequencies with the mean percentage deviations of -1.3, -0.9, -0.9, and +2.2 % for
44 410 SiH4, Si2H6, GeH4, and Ge2H6, respectively.
45
46 411
47
48 412 Thermochemical Properties
49
50
413 The thermodynamic properties of 7 acyclic Si, Ge, and SiGe hydrides with experimental data
51 414 available were estimated using the three different quantum chemical methods mentioned in the
52
53 415 Computational Methodology section. The deviations between the calculated and experimental
54 416 values for standard enthalpy of formation, standard entropy, and constant pressure heat capacity
55
56 13 | P a g e
57
58
59
1
2
3 417 are listed in Table 2 for these acyclic species. Among the methods used, the G3//B3LYP method
4
5 418 was the most accurate for the standard enthalpy of formation calculation for silicon hydrides with
6
419 an average absolute deviation of 5.4 kJ mol-1 from experimental measurements while the
7
8 420 G4//B3LYP method was more accurate for the prediction standard enthalpy of formation values
9
10 421 for germanium hydrides with an average absolute deviation of 19.9 kJ mol-1 from experimental
11 422 measurements. The G3//B3LYP method had similar predictive accuracy as the G4//B3LYP
12
13 423 method for the standard enthalpy of formation for germanium hydrides with an average absolute
14 424 deviation of 22.4 kJ mol-1. The largest deviations between experimental and calculated predictions
15
16 425 for standard enthalpy of formation values observed for germanium hydrides can be attributed to
17
426 factors not included in the composite methods discussed. This conclusion is supported by
18
19 427 reasonably accurate predictions of geometry parameters and vibrational frequencies for the
20
21 428 germanium hydrides discussed in the Structures and Vibrational Frequencies section. The
22 429 quantum chemical factors that lead to large deviations in standard enthalpy of formation prediction
23
24 430 include correlation of core and core-valence electrons and relativistic effects such as spin-orbit
25
431 coupling which becomes progressively more important as heavier elements like Ge are
26
27 432 considered. It should be noted that the G3//B3LYP method does include an experimental spin-
28
29 433 orbit energy correction term used for atoms and calculated spin-orbit energy correction term for
30 434 selected diatomic species.44 For the G3//B3LYP method, the use of a higher level correction factor
31
32 435 based upon a regression of correction parameters from an experimental data set containing
33 436 standard enthalpies of formation, ionization potentials, electron affinities, and proton affinities
34
35 437 differs for Si and Ge hydride species. Namely, there are more Si-based species than Ge-based
36
438 species in this test set for regression of higher level correction factors, i.e., the test set included
37
38 439 species with one or two Si atoms and only one Ge atom.
39
40 440
41 441 To the best of our knowledge, there is no extensive investigation of thermochemical property
42
43 442 estimation or even of vibrational frequency calculations for optimized hydrogenated Si, Ge, and
44 443 SiGe clusters outside of the works cited in this paper. A total of 46 molecules up to a moderate
45
46 444 cluster size, where (Si + Ge) ≤ 6, were investigated in this study. The thermochemical properties
47
48
445 of all 46 species studied here are reported in the Table 4. The most stable ground electronic state
49 446 for all molecules in this study was found to be the singlet state. As mentioned in the previous
50
51 447 section Structures and Vibrational Frequencies, a significant conformational change or
52 448 spontaneous bond dissociation was observed during optimization of the electronic wavefunction
53
54 449 to the triplet state. In all geometries, the addition of Ge atoms to a species increases the standard
55
56 14 | P a g e
57
58
59
1
2
3 450 enthalpy of formation, standard entropy, and constant pressure heat capacity values. The trend
4
5 451 based on elemental composition for standard enthalpy of formation predictions by the G3//B3LYP
6
452 and G4//B3LYP methods was successfully captured for the acyclic hydrides of this study. Due to
7
8 453 the semiconducting or non-local nature of electron correlation in the larger clusters with cyclic or
9
10 454 polycyclic, rigorous composite methods are required to accurately predict trends in
11 455 thermochemical properties such as standard enthalpy of formation as function of Si and Ge
12
13 456 composition.
14 457
15
16 458 Petersson et al.58 proposed the concept of an isodesmic bond additivity correction (BAC) scheme
17
459 based on the spirit of isodesmic reactions. With this approach, our study compared small acyclic
18
19 460 molecules with experimental data available in order to calculate the bond additivity corrections
20
21 461 necessary for implementation of Eq. 4. The novel BAC parameters for Si, Ge, and SiGe species
22 462 regressed in our study to calculate standard enthalpies of formation are presented in Table 5. The
23
24 463 regression statistics show that the most statistically significant BAC parameter is for the Ge-H
25
464 bond followed by the Ge-Ge bond with p-values of 0.007 and 0.011, respectively. The R2-value
26
27 465 for the full regression was 0.9992 with an F-value of 525.8 and a p-value of 0.002. Thus, the full
28
29 466 regression model is statistically significant at the 99.5% confidence interval. The BAC parameters
30 467 are categorized into five different types according to the Si and/or Ge atoms participating in the
31
32 468 respective bond. For all 46 hydrogenated Si, Ge, and SiGe clusters, the resulting standard
33 469 enthalpies of formation using this BAC approach are listed along with the standard enthalpies of
34
35 470 formation obtained from atomization energies without the BAC approach in Table 4. Figure 2a-2f
36
471 depict parity plots to display trends of the BAC impact on prediction of standard enthalpies of
37
38 472 formation for key geometry series (T, ST, TBP, TP, and Pri). On average, the standard enthalpy
39
40 473 of formation for all clusters in this study increased after implementation of the required BAC
41 474 parameters. The most pronounced BAC effect on the standard enthalpy of formation prediction
42
43 475 was for the clusters comprised of higher Ge atom content than Si atom content, particularly
44 476 because the G3//B3LYP method systematically underestimates the standard enthalpy of
45
46 477 formation of SiGe and Ge hydrides. The absolute difference in standard enthalpies of formation
47
48
478 between the Pri-0 and Pri-6 structures before employing the BAC parameters was 10.8 kJ mol-1.
49 479 After implementation of the BAC parameters, the standard enthalpies of formation of Pri-0 and
50
51 480 Pri-6 structures were 602.6 kJ mol-1 and 677.2 kJ mol-1, respectively. Upon implementation of the
52 481 BAC parameters, the absolute difference in standard enthalpies of formation between the Pri-0
53
54 482 and Pri-6 structures become 74.6 kJ mol-1. Thus, species with more X-X bonds (X = Si or Ge) will
55
56 15 | P a g e
57
58
59
1
2
3 483 exhibit a greater impact of the BAC parameters. This trend in correction of standard enthalpy of
4
5 484 formation values can be clearly seen by observing the parity plots in Figure 2, particularly Figure
6
485 2f for the prismane clusters.
7
8 486
9
10 487 The standard entropies and constant pressure heat capacities for all clusters in this study over
11 488 the temperature range of 298.15 K to 1500 K are presented graphically in Figure 3a-3b and in
12
13 489 Table S1 of the Supporting Information. The average percent deviations between experimental
14 490 and calculated values of standard entropy and constant pressure heat capacity for the seven
15
16 491 acyclic Si, Ge, and SiGe hydrides were +1.3% and +0.6%, respectively, at the G3//B3LYP level
17
492 of theory. It is noteworthy to mention that the external symmetry number was identified for all
18
19 493 clusters to ensure that accurate standard entropy values were predicted. The effect of external
20
21 494 symmetry number on rotational entropy becomes less pronounced at elevated temperatures
22 495 because the aforementioned correction factor, R ln σrot, is not a function of temperature. The
23
24 496 clusters with the highest degree of symmetry were the trigonal planar geometries, particularly the
25
497 pure Si and Ge clusters, TP-0 and TP-4, respectively, which both possess a Td point group
26
27 498 symmetry. The D3h point group symmetry was the next most common point group symmetry
28
29 499 with a high degree of symmetry in this study, particularly for the trigonal bipyrimidal (TBP) and
30 500 prismane (Pri) geometries. Chiral clusters containing at least one Si/Ge atom with four
31
32 501 nonidentical substituents were identified in our study. The presence of one chiral center was
33 502 denoted in Table 4 for four substituted trigonal planar and three prismane clusters. The presence
34
35 503 of a chiral center in a cluster increases the external symmetry number by a factor of two. Figure
36
504 3a-3b display standard entropy and constant pressure heat capacity values, respectively, as a
37
38 505 function of temperature and it is interesting to note that standard entropy values are more sensitive
39
40 506 to temperature variations than constant pressure heat capacity for the clusters in this study. For
41 507 both the trigonal planar and trigonal pyramidal geometries, the range of standard entropy and
42
43 508 constant pressure heat capacity values from 298 K to 1500 K is lower than the range of standard
44 509 entropy and constant pressure heat capacity values over the same temperature range for the
45
46 510 substituted trigonal planar, trigonal bipyramidal, and prismane geometries. This observation is
47
48
511 likely due to the greater number of vibrational degrees of freedom for the larger Si, Ge, and SiGe
49 512 clusters sizes.
50
51 513
52 514 If standard enthalpy of formation values are compared for isomers of a given cluster geometry,
53
54 515 relative stabilities can be identified and ranked accordingly. Cluster isomers are present in this
55
56 16 | P a g e
57
58
59
1
2
3 516 study for the substituted trigonal planar (ST), trigonal bipyramidal (TBP), and prismane (Pri)
4
5 517 geometries. For instance in Table 4 for the trigonal bipyramidal geometry, one can observe that
6
518 the isomer TBP-2a is more stable than isomers TBP-2b and TBP-2c where standard enthalpies
7
8 519 of formation are 303.5 kJ mol-1, 328.4 kJ mol-1, and 336.0 kJ mol-1, respectively. The isomer
9
10 520 TBP-2a differs from TBP-2b and TBP-2c due to the presence of two sp3-hybridized Ge atoms in
11 521 TBP-2a compared to two sp3d2-hybridized Ge atoms in TBP-2b and TBP-2c. A similar trend in
12
13 522 cluster stabilities is also observed for the other TBP isomers in the TBP-1, TBP-3, and TBP-4
14 523 series of cluster geometries, but the differences between the most stable isomer and the least
15
16 524 stable isomer in these geometry series are lower with values of 16.6 kJ mol-1, 21.9 kJ mol-1, and
17
525 6.7 kJ mol-1, respectively. The difference in stability of isomers is similarly pronounced for the
18
19 526 substituted trigonal planar geometry series; however, the greatest difference for the stability of
20
21 527 isomers for this study is in the ST-1 series where the ST-1b isomer is most stable. The ST-1b
22 528 isomer has the molecular formula of Si3Ge1H8 where the Ge is at the center of the cluster and
23
24 529 bound to three Si atoms and one H atom. The differences in stability between the most stable
25
530 isomer and the least stable isomer in the ST-1, ST-2, and ST-3 series are 38.3 kJ mol-1, 20.7 kJ
26
27 531 mol-1, and 19.2 kJ mol-1, respectively. The difference in stability of isomers is least pronounced
28
29 532 for the prismane clusters with differences ranging 10.5 kJ mol-1, 9.2 kJ mol-1, and 10.3 kJ mol-1 for
30 533 the Pri-2, Pri-3, and Pri-4 series, respectively.
31
32 534
33
34 535 Generalization of Electronic Properties and Chemical Stability
35
36 536 The highest occupied and lowest unoccupied molecular orbital energies are very informative
37 537 properties of a molecule or cluster which can be calculated by quantum chemical methods. These
38
39 538 molecular orbitals also assign the electron density as a function of position in the molecule or
40
539 cluster, where electron density for a given molecular orbital i is defined as the square of the
41
42 540 electronic wavefunction, i2. Knowledge of the highest occupied and lowest unoccupied
43
44 541 molecular orbital contours is critical for understanding reactions of clusters as well as
45
542 optoelectronic properties. The foundation of the frontier orbital theory for the prediction of the most
46
47 543 reactive positions in multi-electron systems is based on the highest occupied and lowest
48
49
544 unoccupied molecular orbitals. Reactive molecules or clusters are characterized by a small
50 545 highest occupied molecular orbital-lowest unoccupied molecular orbital (HOMO-LUMO) energy
51
52 546 gap. Both the HOMO and LUMO are the primary molecular orbitals that can be used as predictors
53 547 of chemical stability and optoelectronic properties.
54
55 548
56 17 | P a g e
57
58
59
1
2
3 549 In our computational study, the HOMO–LUMO energy gap is considered to investigate the role of
4
5 550 cluster composition on the chemical stability of hydrogenated Si, Ge, and SiGe clusters. This
6
551 energy gap is a critical parameter which characterizes the chemical reactivity of the hydrogenated
7
8 552 clusters. This chemical reactivity is related to the facility of a molecule to participate in chemical
9
10 553 reactions or to create a novel self-assembled material through non-bonding molecular interactions.
11 554 The HOMO–LUMO energy gap can describe the ability for electrons to move from HOMO to
12
13 555 LUMO and consequently is considered as an important parameter to analyze the chemical
14 556 stability of clusters. For instance, if the HOMO–LUMO energy gap were large for a given Si, Ge,
15
16 557 or SiGe cluster, this value would correspond to a closed shell electronic configuration and high
17
558 chemical stability. On the other hand, smaller HOMO–LUMO energy gaps for given Si, Ge, and
18
19 559 SiGe clusters reflect that the respective cluster may interact easily with other molecules to form a
20
21 560 covalent bond and these types of molecules are on average more chemically reactive.
22 561
23
24 562 Calculated HOMO-LUMO energy gaps of hydrogenated Si and Ge clusters and acyclic Si and Ge
25
563 hydrides at the G3//B3LYP level of theory are presented in Figure 4. The highest HOMO-LUMO
26
27 564 energy gaps are for acyclic Si and Ge hydrides comprised of one or two Si/Ge atoms ranging
28
29 565 from 12.4 to 14.6 eV for Si-containing species and 12.0 to 14.2 eV for Ge-containing species.
30 566 The calculated HOMO-LUMO energy gaps for pure Ge species are always lower than for pure Si
31
32 567 species, where the largest absolute difference of 0.62 eV is observed for the trigonal bipyramidal
33 568 geometry. The smallest absolute difference of 0.06 eV between HOMO-LUMO energy gaps for
34
35 569 pure Si and Ge species is observed for the trigonal pyramidal geometry. The trigonal pyramidal
36
570 geometry is the most strained geometry in our study as previously discussed in the Structures
37
38 571 and Vibrational Frequencies section. On average, the HOMO-LUMO energy gap decreases with
39
40 572 increase in the polycyclic nature of the pure Si or Ge cluster. Figure 5a-5e present calculated
41 573 HOMO-LUMO energy gaps of all clusters in our study, particularly highlighting the effects of alloy
42
43 574 cluster composition and isomers. The most precipitous fall in HOMO-LUMO energy gap is for the
44 575 trigonal bipyramidal geometry followed by the trigonal planar, substituted trigonal planar,
45
46 576 prismane, and trigonal pyramidal geometries at 0.49, 0.44, 0.44, and 0.43 eV, respectively. It is
47
48
577 interesting to note that the corresponding band gap energy of bulk semiconductor materials is
49 578 lowered at elevated temperatures as increased atomic vibrations increase interatomic spacing
50
51 579 which decreases the potential seen by the electrons in the material, thus reducing the size of the
52 580 observed band gap energy. It can be expected that a similar phenomenon may be observed for
53
54 581 the larger clusters in this study. For instance, bulk pure Si has a band gap energy of 1.17 and
55
56 18 | P a g e
57
58
59
1
2
3 582 1.11 eV at 0 and 300 K, respectively, and bulk pure Ge has a band gap energy of 0.744 and 0.660
4
5 583 eV at 0 and 300 K, respectively.72, 73
6
584
7
8 585 Many semiconducting materials or material precursors are characterized as hyperpolarizable and
9
10 586 are analyzed by means of vibrational spectroscopy, i.e. Infrared or Raman spectroscopy. In the
11 587 case of Raman spectroscopy, the corresponding analysis of the electronic wavefunction indicates
12
13 588 that the electron absorption corresponds to the transition from the ground state to the first excited
14 589 state and is conventionally described by the one-electron vertical excitation from the HOMO to
15
16 590 the LUMO. For most clusters in this study, the HOMO is delocalized over the entire structure. By
17
591 contrast, the LUMO is still largely delocalized over the entire structure but also extends well
18
19 592 beyond the center of mass of the nuclei positions. Consequently, the HOMO-LUMO transition
20
21 593 implies an electron density transfer to the limits of the molecular orbitals and this phenomena is
22 594 consistent with semiconducting material behavior. This extreme delocalization of electron density
23
24 595 suggests facile electron density transfer between neighboring clusters in the absence of a formal
25
596 covalent bond formation, which can be useful for the development of self-assembling
26
27 597 nanomaterials. Examples of this LUMO behavior can be seen in Figure S1 of the Supporting
28
29 598 Information for the trigonal planar (T), trigonal pyramidal (TP), and prismane (Pri) geometries,
30 599 where this behavior is most pronounced in structures that contain one or more Ge atoms. Figure
31
32 600 6 presents a comparison of calculated contour surfaces of the frontier molecular orbitals (HOMO,
33 601 LUMO) for the TBP-1 cluster using the B3LYP/6-31g(d), G3//B3LYP, and G4//B3LYP levels of
34
35 602 theory. It is important to highlight how the LUMO contour changes significantly between the
36
603 B3LYP/6-31g(d) and G3//B3LYP levels of theory; however, the HOMO contour is essentially the
37
38 604 same between these two level of theory. The HOMO and LUMO contour predictions at the
39
40 605 G4//B3LYP level of theory show the strongest inclusion of electron correlation effects as can be
41 606 observed in Figure 6 for the TBP-1 cluster.
42
43 607
44 608 Since molecular orbital (MO) theory is by far the most widely used by chemists and chemical
45
46 609 engineers, it is important to place the HOMO-LUMO energy gap in a MO framework for reacting
47
48
610 chemical systems. That is, according to the notation introduced in the Computational Methodology
49 611 section, hard molecules have a large HOMO-LUMO energy gap, and soft molecules have a small
50
51 612 HOMO-LUMO energy gap. A small HOMO-LUMO energy gap is correlated to small vertical
52 613 excitation energies to the manifold of excited energy states. Therefore, soft molecules, with a
53
54 614 smaller energy gap than hard molecules, will be more polarizable by definition. High polarizability
55
56 19 | P a g e
57
58
59
1
2
3 615 is the most characteristic property attributed to soft Si, Ge, and SiGe clusters. HOMO-LUMO
4
5 616 energy gaps should be small for the most favorable bonding or non-bonding interaction between
6
617 molecules or clusters, i.e., both reactants or molecules should exhibit soft character. As listed in
7
8 618 Table 6, the trigonal pyramidal and prismane Si, Ge, and SiGe clusters are the softest species in
9
10 619 this study, and thus most reactive due to smaller energy gaps than the remaining clusters and
11 620 acyclic species. Consequently, these clusters also have a 1:1 Si-to-H or Ge-to-H atomic ratio,
12
13 621 whereas the other species in this study have a lower Si or Ge atomic ratio to H. On average, less
14 622 passivation with hydrogen will result in a more reactive cluster or acyclic species. As defined in
15
16 623 the Computational Methodology section, molecular hardness, softness, and chemical potential
17
624 calculated from the G3//B3LYP level of theory are presented in Table 6.
18
19 625
20
21 626 Developing materials with desired optoelectronic properties has always been at the forefront of
22 627 the semiconducting electronics industry. The optical properties of Si clusters and nanocrystals
23
24 628 have been intensively studied due to the possible technological applications of Si in the
25
629 semiconductor industry. It has been shown that the large HOMO–LUMO energy gaps of metal
26
27 630 encapsulated silicon clusters coupled with their weak reactivity make these structures most
28
29 631 suitable for optical absorption and photoluminescence in the visible region.73 Some of these
30 632 studies have been driven by the desire to understand the quantum effects of confinement in
31
32 633 reduced structural dimensions. To control the triplet/singlet excited states in a designed manner
33 634 for a desired optoelectronic property, the rational adjustment of the singlet-triplet energy gap
34
35 635 (ΔE’ST) between the first singlet (S1) and triplet (T1) excited states is the key as depicted in Figure
36
636 7a-7b. Insight into this latter value can be found by the difference between the HOMO-LUMO
37
38 637 energy gap and the singlet-triplet splitting energy value (ΔEST) between the singlet ground state
39
40 638 (S0) and first excited triplet electronic spin state (T1). Here we complement and extend
41 639 optoelectronic studies for Si, Ge, and SiGe clusters by calculating the HOMO-LUMO energy gap,
42
43 640 which provides insight into the energy gap between the singlet ground state (S0) and first excited
44 641 singlet state (S1), and relating this energy value to the energy splitting between the singlet ground
45
46 642 state (S0) and first excited triplet electronic spin state (T1), (ΔEST). For instance, examination of
47
48
643 Table 1 for the singlet-triplet energy splitting values (ΔEST) of SiH4 and GeH4 reveals values of 3.9
49 644 and 3.6 eV, respectively. Examination of Figure 4 for the HOMO-LUMO energy gaps of SiH4 and
50
51 645 GeH4 reveals values of 14.6 and 14.2 eV, respectively. Similar analysis can be extended to the
52 646 other species in this study using Table 1 for singlet-triplet splitting energy values (ΔEST), and
53
54 647 Figures 4 and 5 for HOMO-LUMO energy gap values. On average, our theoretical studies suggest
55
56 20 | P a g e
57
58
59
1
2
3 648 that the Ge-doped clusters will exhibit a lower HOMO-LUMO energy gap and singlet-triplet
4
5 649 splitting energy value (ΔEST). Our results suggest that the vertical excitation energy from a singlet
6
650 spin state to the excited singlet spin state is positively correlated with the HOMO-LUMO energy
7
8 651 gap for a molecule optimized in the ground singlet spin state. Given our preliminary findings on
9
10 652 varying levels of passivation, our calculations also suggest that other surface termination
11 653 schemes may increase cluster chemical stability such that properties of the respective cluster can
12
13 654 be used for optoelectronic materials. In Figure 7a-7b, representative energy level diagrams of two
14 655 optoelectronic processes determined by singlet-triplet splitting (ΔE’ST) between energies of the
15
16 656 lowest singlet (ES1) and triplet (ET1) excited states are presented for the sake of clarity.
17
657
18
19 658 Rational design and optimization of nanoclusters for semiconducting nanomaterial applications
20
21 659 traditionally requires the systematic synthesis and examination of various cluster molecules. This
22 660 conventional “trial-and-error” approach generally requires considerable time and labor costs.
23
24 661 Prediction of the cluster properties with a machine learning approach would facilitate the rational
25
662 design and optimization of nanomaterials, and this approach would allow the discovery of
26
27 663 nanostructures with desired properties rapidly and efficiently. Developing this “nanostructure
28
29 664 informatics” approach would create a practical method to develop a robust predictive model for
30 665 nanocluster reactivity. A convenient set of quantum chemical parameters to train our models using
31
32 666 a machine learning approach is the fractional electrons transferred during molecular interaction
33 667 of reactants which was previously introduced in the Computational Methodology section. The
34
35 668 fractional electrons transferred during molecular interaction of reactants, N, is correlated to the
36
37 669 degree of nucleophilicity and electrophilicity of the reactants, which is highly desirable for tailored
38 670 nanomaterials design such as self-assembling nanomaterials.
39
40 671
41 672 Implementation of Eq. 7 to calculate fractional electrons transferred during molecular interaction
42
43 673 of reactants, N, using the quantum chemical parameters in Table 6 are presented in Figure 8.
44
45 674 In Figure 8, we present 2809 data points representing N values for the interaction of all 53
46 675 hydrogenated Si, Ge, and SiGe species in this computational study. The graphical representation
47
48 676 in Figure 8 very conveniently visualizes the key nucleophilic and electrophilic characters of all
49
677 potential molecular interactions in this study. Dark red and dark blue regions of this data set
50
51 678 represent interactions with the largest N values. The sign of the N value represents the
52
53 679 directionality of the electron transfer process. For instance, if one chooses a reactant B from the
54 680 x-axis and moves vertically along the y-axis of Figure 8, a negative N value signifies that reactant
55
56 21 | P a g e
57
58
59
1
2
3 681 B is a nucleophile when interacting with the corresponding reactant A for the molecular interaction.
4
5 682 Conversely, reactant A for that same interaction has more electrophilic character during the
6
683 interaction. The N values in Figure 8 are derived directly from quantum chemical calculations at
7
8 684 the G3//B3LYP level of theory; however, our study was generalized beyond these calculations by
9
10 685 introducing a machine learning approach to create a robust multiple linear regression equation to
11 686 predict this N value for molecular interactions not explicitly examined in this study. Eq. 8 is
12
13 687 proposed for this purpose where N is a function of the molecular weight of reactants A and B,
14
15 688 i, the degree of passivation of reactants A and B, I, and the regression coefficients, Ci. The
16 689 degree of passivation of the reactants is defined as the atom count of total heavy atoms, Si and
17
18 690 Ge, divided by the atom count of total hydrogen atoms. The use of the molecular weight and
19
691 degree of cluster passivation allows for the implementation of nanomaterials design efforts
20
21 692 independent of the need to perform computationally expensive quantum chemical calculations
22
23 693 during the initial screening efforts of nanomaterials design.
24 694
25
26 695 ∆𝑁 = 𝐶1 ∗ 𝜔𝐴 + 𝐶2 ∗ 𝜋𝐴 + 𝐶3 ∗ 𝜔𝐵 + 𝐶4 ∗ 𝜋𝐵 Eq. 8
27
28
696
29 697 Table 7a contains two models which follow Eq. 8 and were regressed using a machine learning
30
31 698 approach to statistical data analysis. Table 7b also contains the regression analysis including the
32 699 statistical significance and errors of the different models evaluated. Model 1 is comprised of
33
34 700 regression coefficients, Ci, for a training set of all 2809 data points for N in this study. The overall
35
36
701 model was deemed significant if the F-test satisfied the 99% confidence level (i.e., the p-value
37 702 was below  = 0.01). In fact, the total regression for Model 1 was statistically significant at the
38
39 703 99.99% confidence interval, and all four regression coefficients, Ci, were also statistically
40 704 significant at the 99.99% confidence interval despite the total regression for Model 1 having an
41
42 705 R2-value of 0.6660. Although not presented in Table 7, an additional set of four cross terms were
43
44 706 added to the full regression Eq. 8 of the forms, Cij i j, Cij i j, and Cij i j to determine if the
45 707 R2-value could be improved. These additional cross terms did not improve the R2-value for the
46
47 708 full regression and were not statistically significant at the 95% confidence interval. Figure 9
48
709 displays a parity plot for Model 1 of fractional electrons transferred in eV, N, for the training set
49
50 710 of 2809 molecular interactions calculated from the G3//B3LYP level of theory.
51
52 711
53 712 Finally, the best predictive model for N values and its four regression coefficients, Ci, were
54
55 713 validated using the sensitivity analysis proposed by Mavrovouniotis74. This approach removes 10%
56 22 | P a g e
57
58
59
1
2
3 714 of the molecular interactions randomly (or 281 N values for our study), and the four regression
4
5 715 coefficients are refitted. The new regression coefficients are then used to predict the N values
6
7 716 of the removed molecular interactions. The differences between these N values and the values
8 717 predicted from the original four regression coefficients are then calculated to assess the sensitivity
9
10 718 of the regression coefficients. Model 2 presented in Table 7 evaluates the sensitivity analysis of
11
12
719 the four regression coefficients, Ci, for the prediction of N values. A performance summary for
13 720 the prediction of N values for the removed molecular interactions using the refitted regression
14
15 721 coefficients and the regression coefficients from the full regression containing 2809 molecular
16
722 interactions is presented in Table 7c. The run using the refitted regression coefficients had an
17
18 723 average absolute deviation value of 0.01881 eV for the validation set of 281 molecular interactions
19
20 724 compared to the G3//B3LYP values, which is a negligibly higher error than the errors obtained
21 725 using the regression coefficients from the full regression containing 2809 molecular interactions.
22
23 726 Namely, the average absolute deviation value of 0.01879 eV was obtained for the validation set
24 727 of 281 molecular interactions compared to the G3//B3LYP values when using the regression
25
26 728 coefficients from Model 1. Thus, the predictive capability of our multiple linear regression model
27
28 729 for molecular interactions, or N values, not included in the training set is very good. This
29 730 generalization for the prediction of molecular interaction properties is necessary because
30
31 731 computational and/or experimental investigation of all potential interactions of Si/Ge/SiGe clusters
32 732 is not feasible.
33
34 733
35
734
36
37
38
735 Conclusion
39
736 In summary, DFT calculations were performed to study the relative stabilities, thermodynamic
40
41 737 properties and electronic properties of hydrogenated Si, Ge, and SiGe nanoclusters. For
42
43
738 comparison, the properties of pure SixHy and GexHy clusters are also investigated. The optimized
44 739 geometries of the SixGeyHz clusters were investigated systematically using quantum chemical
45
46 740 calculations and conventional statistical thermodynamics. All electronic energies for the clusters
47 741 were calculated using Gaussian-n methods, which use B3LYP geometries and higher-level
48
49 742 corrections based on single point energies. To validate our approach, we compared our
50
743 computational methodology to other composite methods such as the complete basis set
51
52 744 (CBS-QB3) and G4//B3LYP methods, as well as to available experimental data. The geometry
53
54 745 parameters of all the molecules increased nominally as Ge atoms were substituted for Si atoms;
55
56 23 | P a g e
57
58
59
1
2
3 746 however, the geometric change was small when compared to the changes observed in the
4
5 747 electronic properties. Detailed vibrational frequency analysis has confirmed that all species
6
748 reported in this study are minima on the potential energy surface and possess all real vibrational
7
8 749 frequencies. As Si atoms were exchanged for Ge atoms in a given cluster geometry, the
9
10 750 calculated thermochemical properties increased proportionally with the number of Ge atoms in
11 751 the cluster. The calculated HOMO-LUMO energy gaps are proportionally decreased, as the
12
13 752 cluster size increases in total heavy atom count, Si or Ge atoms.
14 753
15
16 754 Standard enthalpy of formation at 298 K and standard entropy and constant pressure heat
17
755 capacity at elevated temperatures, i.e., 298-1500 K, were calculated for the 46 hydrogenated Si,
18
19 756 Ge, and SiGe clusters and 7 acyclic Si, Ge, and SiGe species in this study using the G3//B3LYP
20
21 757 composite method and statistical thermodynamics with anharmonic vibrational frequency
22 758 corrections. The hydrogenated Si, Ge, and SiGe clusters contained between one and six Si and/or
23
24 759 Ge atoms and polycyclic nature by way of fused three- to four-membered rings, as well as different
25
760 degrees of dehydrogenation or multifunctionality. Quantum chemical descriptors based on the
26
27 761 G3//B3LYP method, electronic chemical potential, , and absolute hardness and softness,  and
28
29 762 , respectively, were calculated and generalized using a machine learning approach to predict
30
763 the reactivity of Si, Ge, and SiGe alloy clusters and acyclic species in the gas phase. A statistically
31
32 764 significant predictive model at the 99.9% confidence interval was regressed to allow for
33
34 765 nanomaterials design efforts independent of the need to perform computationally expensive
35 766 quantum chemical calculations during the initial screening efforts of nanomaterials design.
36
37 767
38
39
768
40
41
769
42
43 770
44
45 771
46
47
48
49
50
51
52
53
54
55
56 24 | P a g e
57
58
59
1
2
3
4 773 Tables and Figures
5
6 774
7 775 Table 1. Calculated singlet-triplet splitting values of selected Si, Ge, and SiGe species using the
8
9 776 G3//B3LYP level of theory, where the splitting value is defined as the difference in total energy of
10 777 the species between the singlet ground state (S0) and the first excited triplet state (T1). The
11
12 778 nomenclature to identify molecular geometries is the same as in Figure 1, and representative
13
779 energy level diagrams can be found in Figure 7. ZPE denotes zero-point vibrational energy.
14
15 780
16
17 　 G3//B3LYP Electronic energies with ZPE correction
18
　　 Spin multiplicity 　　
19
20 　　 Singlet Triplet 　 Singlet-triplet splitting
21 Species Index (Hartrees) (Hartrees) 　 (eV)
22
23 Si1H4 L-1 -291.7112 -291.5693 a 3.9
24 Ge1H4 L-5 -2078.819 -2078.6874 b 3.6
25 Si3H6 T-0 -871.5746 -871.5151 　 1.6
26
27 Ge3H6 T-3 -6232.9354 -6232.8851 　 1.4
28 Si5H8 TBP-0 -1451.5169 -1451.4273 　 2.4
29
Si4GeH8 TBP-1a -3238.6407 -3238.5581 　 2.2
30
31 Si3Ge2H8 TBP-2b -5025.7602 -5025.6793 　 2.2
32 Si4Ge2H6 Pri-2b -5313.8816 -5313.8201 　 1.7
33
34 Si3Ge3H6 Pri-3c -7101.0080 -7100.9437 　 1.7
　　　　　　
35
36 a This ·
molecule exists in a form of dissociation of SiH3 and H ·
37
38 b This molecule exists in a form of dissociation of ·GeH 3 and ·H
39 781
40
41 782
42 783
43
44 784
45
785
46
47 786
48
49
787
50 788
51
52 789
53 790
54
55 791
56 25 | P a g e
57
58
59
1
2
3 792 Table 2. Comparison of calculated standard enthalpy of formation, standard entropy, and constant
4
5 793 pressure heat capacity at 298 K to available experimental data for small acyclic hydrogenated
6
794 silicon and germanium species using the CBS-QB3, G3//B3LYP, and G4//B3LYP composite
7
8 795 methods. Deviation is defined as experiment minus theory. AAD denotes average absolute
9
10 796 deviation.
11 ◦
net spin
12 Molecules charge multiplicity ∆H f,298K
(kJ mol-1)
13
a
14 level of theory Exp CBS-QB3 Deviation G3B3 Deviation G4 Deviation
15 Si1H4 01 34.3 22.57 11.7 31.22 3.1 31.52 2.8

16 Si2H6 01 80.3 65.39 14.9 75.69 4.6 75.03 5.3
17 Si3H8 01 120.9 100.38 112.46 111.49
20.5 8.4 9.4
18 AAD 　　 15.7 5.4 5.8
19 SiGeH6 01 116.3 71.70
44.6
107.26
9.1
108.62 7.7
20 Ge1H4 9.3
01 90.8 47.48 43.3 80.71 10.1 81.46
21 Ge2H6 20.1
01 162.3 79.07 83.2 139.71 22.6 142.23
22 Ge3H8 30.3
01 226.8 108.68 118.1 192.40 34.4 196.51
23 AAD 81.6 22.4 19.9
24 298K
　　　 S
25 (J mol-1 K-1)
　　　
26
27 　　　 Exp
a
CBS-QB3 Deviation G3B3 Deviation G4 Deviation
28 Si1H4 01 204.6 204.38 0.2 204.56 0.0 204.48 0.1

29 Si2H6 01 272.7 273.82 -1.1 273.57 -0.9 274.10 -1.4
30
Si3H8 01 N.A. 346.58 347.08 347.38
31 AAD
-
0.7
-
0.5
-
0.8
　　　　　　
32
SiGeH6 01 N.A.
33 292.66
-
294.27
-
292.91
-
34 Ge1H4 01 217.1 217.31 -0.2 217.48 -0.4 217.41 -0.3

35 Ge2H6 01 297.0 300.47 -3.5 304.20 -7.2 300.88 -3.9
36 Ge3H8 01 N.A. 388.09 - 419.08 - 389.24 -
37 AAD 　　　　 1.9 　 3.8 　 2.1
38 　　　
Cp
39 　　　 (J mol-1 K-1)
40 a
Exp CBS-QB3 Deviation G3B3 Deviation G4 Deviation
41 　　　
42 Si1H4 01 42.8 42.52 0.3 42.71 0.1 42.70 0.1
43 Si2H6 01 80.0 79.73 0.3 79.49 0.5 79.85 0.2
44 Si3H8 01 N.A. 117.55 - 117.13 - 117.71 -
45 AAD 　　　　 0.3 　 0.3 　 0.2
46 SiGeH6 01 N.A. 82.37 - 82.95 - 82.65 -
47
Ge1H4
48 01 45.0 44.78 0.2 44.88 0.1 45.06 -0.1
Ge2H6
49 01 84.9 85.16 -0.3 86.15 -1.2 85.60 -0.7
50 Ge3H8 01 N.A. 126.17 - 128.35 - 126.92 -

AAD 0.3 1.1 0.4
51 　　　　　
52
53 797
54 798 Table 3. Comparison of experimental vibrational modes for SiH4, Si2H6, GeH4, and Ge2H6 to
55
56 26 | P a g e
57
58
59
1
2
3 799 unscaled harmonic vibrational modes using the G3//B3LYP and G4//B3LYP composite methods.
4
5 800 All vibrational frequency values are reported in cm-1.
6 (a) Vibrational Modes of SiH4 　　　　　
7
G3//B3LYP G4//B3LYP
8 experimental
mode calc. deviation % calc. deviation %
9 symmetry frequencya (exp- (exp-
10 freq. deviation freq. deviation
calc) calc)
11 A1 2187 2252 -65 -3.0 2238 -51 -2.3
12 E 975 975 0 0.0 975 0.2 0.0
13 T2 2191 2265 -74 -3.4 2248 -56.6 -2.6
14
T2 914 917 -3 -0.3 918 -4.3 -0.5
15
16 av 　　 -35.5 -1.7 　 -27.9 -1.3
17
(b) Vibrational Modes of Si2H6　
18
19 G3//B3LYP G4//B3LYP
mode experimental calc.
deviation % calc. deviation %
20
symmetry frequencyb (exp- (exp-
21 freq. deviation freq. deviation
calc) calc)
22
A1g 2152 2239 -86.9 -4.0 2217 -65.0 -3.0
23
A1g 909 930 -21.3 -2.3 926 -17.2 -1.9
24
25 A1g 434 433 0.8 0.2 422 11.7 2.7
26 A1u 131 127 3.6 2.7 127 4.4 3.4
27 A2u 2154 2229 -75.1 -3.5 2210 -55.9 -2.6
28 A2u 844 855 -11.5 -1.4 853 -9.0 -1.1
29 Eg 2155 2239 -83.9 -3.9 2219 -64.0 -3.0
30 Eg 929 943 -13.8 -1.5 944 -14.6 -1.6
31 Eg 625 638 -12.7 -2.0 632 -7.3 -1.2
32
Eu 2179 2252 -72.6 -3.3 2228 -49.0 -2.2
33
Eu 940 957 -17.2 -1.8 957 -17.2 -1.8
34
35 Eu 379 380 -0.6 -0.2 373 6.4 1.7
36 av 　　 -25.8 -1.8 　 -16.8 -0.9
37
38 (c) Vibrational Modes of GeH4　
40 mode experimental calc.
deviation % calc. deviation %
41 symmetry frequencyc (exp- (exp-
freq. deviation freq. deviation
42 calc) calc)
43 A1 2106 2051 54.7 2.6 2153 -46.7 -2.2
44 E 931 936 -4.9 -0.5 924 6.9 0.7
45 T2 2114 2082 32.5 1.5 2160 -45.9 -2.2
46 T2 819 824 -5.0 -0.6 819 0.0 0.0
47 av 　　 19.3 0.7 　 -21.4 -0.9
48
49 (d) Vibrational Modes of Ge2H6　
51 mode experimental calc. deviation % calc. deviation %
52 symmetry frequencyd (exp- (exp-
freq. deviation freq. deviation
53 calc) calc)
54 A1g 2068 2025 42.8 2.1 2124 -55.9 -2.7
55
56 27 | P a g e
57
58
59
1
2
3 A1g 832 835 -2.6 -0.3 838 -5.6 -0.7
4 A1g 268 238 29.9 11.2 254 14.2 5.3
5 A1u 146 80 66.0 45.2 105 40.8 28.0
6
A2u 2077 2031 45.8 2.2 2130 -52.9 -2.5
7
8 A2u 756 752 4.4 0.6 753 2.8 0.4
9 Eu 2091 2067 24.4 1.2 2142 -51.4 -2.5
10 Eu 879 887 -7.9 -0.9 887 -7.7 -0.9
11 Eu 370 349 20.7 5.6 354 15.7 4.2
12 Eg 2081 2058 22.9 1.1 2134 -52.7 -2.5
13 Eg 880 891 -10.7 -1.2 892 -11.8 -1.3
14 Eg 567 548 19.4 3.4 557 10.2 1.8
15
av 　　 21.3 5.8 　 -12.9 2.2
16
17 a,c ref75, b ref76
18
19
20 801
21
802
22
23 803
24
25 804
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 28 | P a g e
57
58
59
1
2
3 805 Table 4. Comparison of calculated thermodynamic properties of hydrogenated Si, Ge, and SiGe
4
5 806 clusters using the G3//B3LYP method with and without the bond-additivity corrections (BAC) as
6
807 denoted in Eq. 4. The nomenclature to identify molecular geometries is the same as in Figure 1.
7
8 Level of Theory 　 G3//B3LYP G3//B3LYP (BAC)
9 　Atoms Spin Symmetry ∆H f,298K
◦ Cp S ∆H f,298K
◦ Cp S
10 Net multi
11
　Si Ge H Species charge -plicity group chirality sext KJ/mol J/mol.K J/mol.K KJ/mol J/mol.K J/mol.K
12
13 Trigonal Planar 　　　　　　　　　　　
14 　3 0 6 T-0 0 1 D3h 　 6 261.1 105.4 319.5 268.9 105.3 304.5
15 　2 1 6 T-1 0 1 C2v 　 2 278.8 110.3 339.1 286.4 110.2 333.3
16 　1 2 6 T-2 0 1 C2v 　 2 296.1 114.7 358.8 313.2 114.6 352.8
17
　0 3 6 T-3 0 1 D3h 　 6 313.8 118.7 378.2 349.9 118.6 363.1
18
19 Trigonal Pyramidal 　　　　　
20 　4 0 4 TP-0 0 1 Td 　 12 644.0 122.1 352.9 654.1 121.9 331.9
21 　3 1 4 TP-1 0 1 C3v 　 3 655.2 122.7 359.6 661.4 122.6 350.2
22 　2 2 4 TP-2 0 1 C2v 　 2 665.7 123.4 367.7 677.5 123.3 361.9
23
　1 3 4 TP-3 0 1 C3v 　 3 675.1 124.3 376.1 702.4 124.2 366.9
24
25 　0 4 4 TP-4 0 1 Td 　 12 683.5 125.4 384.6 735.7 125.4 363.8
26 Substituted trigonal planar
27 　4 0 8 ST-0 0 1 CS 　 1 282.7 142.0 381.5 293.3 141.9 381.3
28 　3 1 8 ST-1a 0 1 CS 　 1 316.1 145.8 402.6 330.3 145.7 402.4
29
　3 1 8 ST-1b 0 1 CS 　 1 285.4 146.2 394.5 292.0 146.1 394.3
30
31 　3 1 8 ST-1c 0 1 C1 o 2 300.8 147.0 402.5 311.1 146.9 402.4
32 　2 2 8 ST-2a 0 1 CS 　 1 320.0 149.6 417.7 339.8 149.5 417.5
33 　2 2 8 ST-2b 0 1 C1 o 2 334.2 150.8 425.3 348.2 150.7 425.1
34 　2 2 8 ST-2c 0 1 C1 o 2 303.1 150.7 413.0 319.1 150.5 412.8
35
　2 2 8 ST-2d 0 1 CS 　 1 318.5 151.6 424.0 338.3 151.5 423.8
36
37 　1 3 8 ST-3a 0 1 C1 o 2 337.5 154.0 437.4 366.8 153.9 437.2
38 　1 3 8 ST-3b 0 1 CS 　 1 351.9 155.3 451.4 375.3 155.2 451.2
39 　1 3 8 ST-3c 0 1 CS 　 1 321.0 154.9 431.8 356.1 154.8 431.6
40 　0 4 8 ST-4 0 1 CS 　 1 355.2 158.2 464.2 403.6 158.1 464.0
41
Trigonal Bipyramidal 　　　　　
42
43 　5 0 8 TBP-0 0 1 D3h 　 6 282.4 157.6 373.9 295.7 157.4 358.8
44 　4 1 8 TBP-1a 0 1 C3v 　 3 290.0 160.3 385.7 299.4 160.1 376.3
45 　4 1 8 TBP-1b 0 1 C2v 　 2 302.9 162.9 392.0 316.0 162.8 386.2
46 　3 2 8 TBP-2a 0 1 D3h 　 6 298.1 162.4 396.3 303.5 162.2 381.1
47
　3 2 8 TBP-2b 0 1 CS 　 1 309.6 165.4 403.3 328.4 165.2 403.1
48
49 　3 2 8 TBP-2c 0 1 C2v 　 2 323.2 168.4 411.3 336.0 168.3 405.4
50 　2 3 8 TBP-3a 0 1 C2v 　 2 317.0 166.8 412.5 335.3 166.6 406.5
51 　2 3 8 TBP-3b 0 1 D3h 　 6 343.4 173.8 432.3 356.0 173.7 417.2
52 　2 3 8 TBP-3c 0 1 CS 　 1 329.1 170.4 422.0 357.2 170.3 421.9
53
　1 4 8 TBP-4a 0 1 C3v 　 3 348.4 175.5 442.4 386.0 175.4 433.1
54
55
56 29 | P a g e
57
58
59
1
2
3 　1 4 8 TBP-4b 0 1 C2v 　 2 335.7 171.4 430.0 379.3 171.2 424.1
4 　0 5 8 TBP-5 0 1 D3h 　 6 354.2 176.0 448.8 416.8 175.8 433.9
5
Prismane 　　　　　　　
6
7 　6 0 6 Pri-0 0 1 D3h 　 6 587.1 171.3 393.8 602.6 171.2 378.7
8 　5 1 6 Pri-1 0 1 CS 　 1 589.5 174.0 405.7 601.0 173.8 405.5
9 　4 2 6 Pri-2a 0 1 CS 　 1 591.6 176.1 416.7 608.8 176.0 416.5
10 　4 2 6 Pri-2b 0 1 C2v 　 2 592.7 175.4 415.6 609.9 175.3 409.7
11
　4 2 6 Pri-2c 0 1 C2 o 4 591.9 176.4 417.8 599.4 176.3 411.8
12
13 　3 3 6 Pri-3a 0 1 C3v 　 3 593.5 178.0 427.2 626.0 177.9 417.9
14 　3 3 6 Pri-3b 0 1 C1 o 2 594.2 177.7 427.2 617.1 177.6 427.0
15 　3 3 6 Pri-3c 0 1 CS 　 1 593.9 178.4 428.9 616.8 178.2 428.7
16 　2 4 6 Pri-4a 0 1 CS 　 1 595.7 179.7 438.2 633.9 179.6 437.9
17
　2 4 6 Pri-4b 0 1 C2v 　 2 596.5 178.9 436.9 634.7 178.8 430.9
18
19 　2 4 6 Pri-4c 0 1 C2 o 4 595.8 179.7 438.8 624.4 179.6 432.8
20 　1 5 6 Pri-5 0 1 CS 　 1 597.3 181.1 448.6 650.9 181.0 448.3
21 　0 6 6 Pri-6 0 1 D3h 　 6 598.7 182.7 459.0 677.2 182.6 443.9
22
808
23
24 809
25
26 810
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 30 | P a g e
57
58
59
1
2
3 811 Table 5. Summary of regressed parameters for the Bond Additivity Correction (BAC) of different
4
812 bond types for standard enthalpy of formation at 298 K calculated from atomization energies and
5
6
813 the G3//B3LYP level of theory.
7
　 Si-H Ge-H Si-Ge Si-Si Ge-Ge
8
9 BAC values
-0.12 1.83 -0.66 1.28 7.06
10 (KJ/mol)
11 814
12
13
815
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 31 | P a g e
57
58
59
1
2
3 Table 6. Calculated quantum chemical molecular descriptors for hardness (, chemical potential
4
5
(, and softness ( at the G3//B3LYP level of theory for all hydrogenated Si, Ge, and SiGe
6 clusters and acyclic species in this study. Hardness in eV, chemical potential in eV, and softness
7 in eV-1.
8
9
Index   
10
11 1 L-1 7.3 -5.9 0.14
12 2 L-2 6.2 -4.9 0.16
13 3 L-3 5.9 -4.6 0.17
14
15 4 L-4 6.1 -4.8 0.16
16 5 L-5 7.1 -5.7 0.14
17 6 L-6 6.0 -4.7 0.17
18
19
7 L-7 5.7 -4.4 0.18
20 8 T-0 5.2 -3.8 0.19
21 9 T-1 5.1 -3.7 0.20
22
23
10 T-2 5.0 -3.7 0.20
24 11 T-3 5.0 -3.7 0.20
25 12 TP-0 4.1 -2.9 0.24
26 13 TP-1 4.0 -2.9 0.25
27
28 14 TP-2 3.9 -3.0 0.26
29 15 TP-3 3.8 -2.9 0.26
30 16 TP-4 3.8 -2.9 0.26
31
32
17 ST-0 5.1 -3.8 0.20
33 18 ST-1a 5.1 -3.8 0.20
34 19 ST-1b 5.0 -3.7 0.20
35
36
20 ST-1c 5.0 -3.7 0.20
37 21 ST-2a 5.0 -3.7 0.20
38 22 ST-2b 5.0 -3.7 0.20
39 23 ST-2c 4.9 -3.6 0.20
40
41 24 ST-2d 4.9 -3.6 0.20
42 25 ST-3a 4.9 -3.6 0.20
43 26 ST-3b 4.9 -3.6 0.21
44
45
27 ST-3c 4.9 -3.6 0.20
46 28 ST-4 4.9 -3.6 0.21
47 29 TBP-0 5.2 -3.9 0.19
48
49
30 TBP-1a 5.1 -3.9 0.20
50 31 TBP-1b 5.0 -3.8 0.20
51 32 TBP-2a 5.0 -3.9 0.20
52 33 TBP-2b 4.9 -3.8 0.20
53
54 34 TBP-2c 5.0 -3.8 0.20
55
56 32 | P a g e
57
58
59
1
2
3 35 TBP-3a 4.9 -3.8 0.21
4
5 36 TBP-3b 5.0 -3.9 0.20
6 37 TBP-3c 4.9 -3.8 0.20
7 38 TBP-4a 4.9 -3.8 0.20
8
9 39 TBP-4b 4.8 -3.8 0.21
10 40 TBP-5 4.8 -3.8 0.21
11 41 Pri-0 4.6 -3.4 0.22
12
13
42 Pri-1 4.5 -3.4 0.22
14 43 Pri-2a 4.4 -3.4 0.22
15 44 Pri-2b 4.4 -3.4 0.22
16 45 Pri-2c 4.5 -3.4 0.22
17
18 46 Pri-3a 4.4 -3.4 0.22
19 47 Pri-3b 4.4 -3.4 0.23
20 48 Pri-3c 4.4 -3.4 0.23
21
22 49 Pri-4a 4.4 -3.4 0.23
23 50 Pri-4b 4.4 -3.4 0.23
24 51 Pri-4c 4.4 -3.4 0.23
25
26
52 Pri-5 4.4 -3.4 0.23
27 53 Pri-6 4.4 -3.4 0.23
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 33 | P a g e
57
58
59
1
2
3 Table 7. (a) Coefficients for the full regression containing 2809 N values (Model 1) and for the regression
4 containing 10% of the N values randomly removed (Model 2), (b) statistical analysis for the least squares
5 regressions and summary of errors, (c) performance summary of the sensitivity analysis using the refitted
6 coefficients (Model 2) and the coefficients from the full regression (Model 1). AAD denotes average absolute
7 deviation.
8
9 (a)
10 Regression coefficients
11
C1 C2 C3 C4
12
13 Model 1 0.08245 0.000071 -0.08245 -0.000071
14 Model 2 0.08309 0.000067 -0.08186 -0.000074
15
16
17
18
(b)
19
20 Training set
21 Regression F-Test
22
23 R2-value F-value P-value AAD (eV) std dev
24 Model 1 0.6660 1398.46 <0.001 0.0172 0.0144
25 Model 2 0.6690 1275.56 <0.001 0.0170 0.0143
26
27
28
29 (c)
30
31 Validation set
32 Regression
33 parameters AAD (eV) std dev
34
35 Model 1 0.01879 0.01491
36 Model 2 0.01881 0.01491
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 34 | P a g e
57
58
59
1
2
3 Figure 1. Optimized SixGeyHz (x+y=6) cluster geometries using the G3//B3LYP level of theory.
4
5 The clusters are denoted by T for trigonal planar, TP for trigonal pyramidal, ST for substituted
6
trigonal planar, TBP for trigonal bipyramidal, and Pri for prismane geometries. The indices are
7
8 incremented by integer values to correspond with the replacement of a Si (yellow) atom by a Ge
9
10 (green) atom from 0 to N, where 0 is the pure Si cluster and N is the pure Ge cluster. The lower
11 case letter symbol denotes isomers.
12
13
14
15
16
17 T-0 T-1 T-2 T-3 TP-0 TP-1 TP-2
18
19
20
21
22
23 TP-3 TP-4 ST-0 ST-1a ST-1b ST-1c ST-2a
24
25
26
27
28
29 ST-2b ST-2c ST-2d ST-3a ST-3b ST-3c ST-4
30
31
32
33
34
35
36 TBP-0 TBP-1a TBP-1b TBP-2a TBP-2b TBP-2c TBP-3a
37
38
39
40
41
42
TBP-3b TBP-3c TBP-4a TBP-4b TBP-5 Pri-0 Pri-1
43
44
45
46
47
48
49 Pri-2a Pri-2b Pri-2c Pri-3a Pri-3b Pri-3c Pri-4a
50
51
52
53
54
55
Pri-4b Pri-4c Pri-5 Pri-6
56 35 | P a g e
57
58
59
1
2
3 Figure 2. Parity plots of standard enthalpy of formation for the 46 hydrogenated Si, Ge, and SiGe
4
5 clusters in this study: (a) all geometries, (b) trigonal planar group, (c) substituted trigonal planar
6
group, (d) trigonal bipyramidal group, (e) trigonal pyramidal group, and (f) prismane group.
7
8
Heat of Formation with BAC (KJ/mol)

9 800 400
10 (a) (b)
11 700
12 350
13 600
14 500 300
15 T
16 400 TP
17 ST 250
18 300 TBP T
19 Pri
200 200
20 200 300 400 500 600 700 800 200 250 300 350 400
21 Heat of Formation (KJ/mol) Heat of Formation (KJ/mol)
22
23

450 450
24
(c) (d)
25
26 400 400
27
28
29 350 350
30
31 300
300
32
33 ST TBP
34 250 250
35 250 300 350 400 450 250 300 350 400 450
37
38
39 750 750
40 (e) (f)
41 700 700
42
43
44 650 650
45
46
600 600
47
48 TP Pri
49 550 550
50 550 600 650 700 750 550 600 650 700 750
52
53
54
55
56 36 | P a g e
57
58
59
1
2
3 Figure 3. (a) Standard entropies and (b) constant pressure heat capacities of all hydrogenated
4
Si, Ge, and SiGe clusters in this study over the temperature range of 298.15K to 1500 K using
5
6
the G3//B3LYP level of theory.
7
8 900
1500K
(a)
Standard Entropy [J/mol.K]
9 800 1400K
10 1300K
700 1200K
11
12 600 1100K
1000K
13 500 900K
14 800K
400
15 600K
16 300 500K
17 400K
200 300K
18 Linear T TP ST TBP Pri 298.15K
100
19
20
21 300 (b) 1500K
Heat Capacity [J/mol.K]
22 1400K
23 250 1300K
24 1200K
200 1100K
25 1000K
26 150 900K
27 800K
28 100 600K
500K
29
50 400K
30 300K
31 Linear T TP ST TBP Pri 298.15K
0
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 37 | P a g e
57
58
59
1
2
3 Figure 4. Comparison of calculated HOMO-LUMO energy gaps for pure silicon and germanium
4
clusters using the G3//B3LYP level of theory. The nomenclature to identify cluster geometries is
5
6
the same as in Figure 1.
7
8
9 HOMO-LUMO Energy of Si and Ge clusters
10 15
11 Silicon clusters
14
12
HOMO-LUMO gap (eV)
Germanium clusters
13 13
14 12
15
16 11
17 10
18
19 9
20 8
21
7
22
23 6
24 X1 X2 X3 H X3H6 X4H8( X4H4( X5H8(T X6H6(Pr
H4 H6 8(L) (T) ST) TP) BP) i)
25
26 Increasing the number of Si (Ge) atoms
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 38 | P a g e
57
58
59
1
2
3 Figure 5. Comparison of calculated HOMO-LUMO energy gaps for all hydrogenated Si, SiGe, and
4
5 Ge clusters in this study using the G3//B3LYP level of theory. The nomenclature to identify cluster
6
geometries is the same as in Figure 1.
7
8
9 (a) Trigonal planar (b) Trigonal pyramidal
11
10
HOMO-LUMO gap (eV)
11
10
12
13
14 9
15
16 8
17
18 7
T-0 T-1 T-2 T-3 TP-0 TP-1 TP-2 TP-3 TP-4
19
20 (c) Substituted trigonal planar
HOMO-LUMO gap (eV)
21 11
22
10
23
24 9
25
26 8
27
28 7
29 ST-0 1a 1b 1c 2a 2b 2c 2d 3a 3b 3c 4
30
(d) Trigonal Bipyramidal
31 11
HOMO-LUMO gap (eV)
32
33 10
34
35 9
36
37 8
38
39 7
40 TBP-0 1a 1b 2a 2b 2c 3a 3b 3c 4a 4b 5
41 (e) Prismane
42 11
HOMO-LUMO gap (eV)
43
44 10
45
46 9
47
48 8
49
50 7
51
Pri-0 1 2a 2b 2c 3a 3b 3c 4a 4b 4c 5 6
52 Increasing the number of Ge atoms
53
54
55
56 39 | P a g e
57
58
59
1
2
3 Figure 6. Comparison of calculated contour surfaces of frontier molecular orbitals (HOMO, LUMO)
4
5 for the TBP-1 cluster using the B3LYP/6-31g(d), G3//B3LYP, and G4//B3LYP levels of theory.
6
The HOMO and LUMO orbital distributions are presented using an isovalue of 0.02. The
7
8 nomenclature to identify cluster geometry is the same as in Figure 1.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 40 | P a g e
57
58
59
1
2
3 Figure 7. Representative energy level diagrams of two optoelectronic processes determined by
4
5 singlet-triplet splitting (ΔE’ST) between energies of the lowest singlet (ES1) and triplet (ET1) excited
6
states. Process (a) has a small ΔE’ST value, and process (b) has a large ΔE’ST value. The
7
8 singlet-triplet splitting values (ΔEST) between energy of the ground state singlet (ES0) and first
9
10 excited triplet state (ET1) which are reported in Table 1 for selected species are also labelled for
11 the sake of clarity. Downward arrows are associated with emissions, and upward arrows are
12
13 associated with transitions.
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 41 | P a g e
57
58
59
1
2
3 Figure 8. Contour map of the calculated fractional electrons transferred in eV (N) for molecular
4
5 interactions of all 53 molecules in this study. N follows Eq. 3 where the frontier molecular orbital
6
7 energies are calculated using the G3//B3LYP level of theory. The reactant index number follows
8 the numbering scheme of Table 6.
9
10
11
12 Fractional number of electrons transferred, N (eV)
13
14 50
15
16 0.1
45
17
18 40
Reactant A Index Number
19
20 0.05
35
21
22
30
23
24 0
25
25
26
20
27
28 -0.05
15
29
30
10
31
-0.1
32
33 5
34
35 5 10 15 20 25 30 35 40 45 50
36
Reactant B Index Number
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 42 | P a g e
57
58
59
1
2
3 Figure 9. Parity plot of fractional electrons transferred in eV for the training set of 2809 molecular
4
5 interactions from the G3//B3LYP level of theory. ML denotes prediction of fractional electrons
6
transferred using the machine learning model regressed in this study.
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 43 | P a g e
57
58
59
1
2
3
4 Supporting Information
5
6 Calculated standard entropies and constant pressure heat capacities at elevated temperatures
7 using the G3//B3LYP method for all hydrogenated Si, Ge, and SiGe clusters and acyclic species
8
9 in this study (Table S1). Calculated contour surfaces of frontier molecular orbitals (HOMO, LUMO)
10 for all hydrogenated Si, Ge, and SiGe clusters in this study using the G3//B3LYP level of theory
11
12 (Figure S1). This material is available free of charge via the Internet at http://pubs.acs.org.
13
14
15 Acknowledgements
16
17 We are grateful for the support of this work by the Alabama Supercomputing Center, Auburn
18 University Hopper high-performance compute cluster resources, and Auburn University new
19
20 faculty start-up funding.
21
22
23
24 References
25
26 1. Harame, D. L.; Koester, S. J.; Freeman, G.; Cottrel, P.; Rim, K.; Dehlinger, G.; Ahlgren, D.; Dunn, J.
27 S.; Greenberg, D.; Joseph, A. at el. The Revolution in Sige: Impact on Device Electronics. Appl. Surf. Sci.
28 2004, 224, 9-17.
29 2. Pham, D. P.; Kim, S.; Park, J.; Le, A. H. T.; Cho, J.; Jung, J.; Iftiquar, S. M.; Yi, J. Reduction in
30 Photocurrent Loss and Improvement in Performance of Single Junction Solar Cell Due to Multistep Grading
31 of Hydrogenated Amorphous Silicon Germanium Active Layer. Silicon 2018, 10, 759-767.
32 3. Ji, X. Y.; Cheng, H. Y.; Grede, A. J.; Molina, A.; Talreja, D.; Mohney, S. E.; Giebink, N. C.; Badding,
33 J. V.; Gopalan, V. Conformal Coating of Amorphous Silicon and Germanium by High Pressure Chemical
34 Vapor Deposition for Photovoltaic Fabrics. APL Mater. 2018, 6, 046105.
35 4. Leitz, C. W.; Currie, M. T.; Lee, M. L.; Cheng, Z. Y.; Antoniadis, D. A.; Fitzgerald, E. A. Hole
36 Mobility Enhancements and Alloy Scattering-Limited Mobility in Tensile Strained Si/Sige Surface Channel
37
Metal-Oxide-Semiconductor Field-effect Transistors. J. App. Phys. 2002, 92, 3745-3751.
38
5. Yoffe, A. D. Semiconductor Quantum Dots and Related Systems: Electronic, Optical, Luminescence
39
and Related Properties of Low Dimensional Systems. Adv. Phys. 2001, 50, 1-208.
40
6. Paul, D. J. Silicon-Germanium Strained Layer Materials in Microelectronics. Adv. Mater. 1999, 11,
41
42
191-204.
43
7. Adamczyk, A. J.; Reyniers, M. F.; Marin, G. B.; Broadbelt, L. J. Kinetics of Substituted Silylene
44 Addition and Elimination in Silicon Nanocluster Growth Captured by Group Additivity. ChemPhysChem 2010,
45 11, 1978-1994.
46 8. Adamczyk, A. J.; Broadbelt, L. J. Thermochemical Property Estimation of Hydrogenated Silicon
47 Clusters. J. Phys. Chem. A 2011, 115, 8969-8982.
48 9. Adamczyk, A. J.; Broadbelt, L. J. The Role of Multifunctional Kinetics during Early-Stage Silicon
49 Hydride Pyrolysis: Reactivity of Si(2)H(2) Isomers with SiH(4) and Si(2)H(6). J. Phys. Chem. A 2011, 115,
50 2409-2422.
51 10. Adamczyk, A. J.; Reyniers, M.-F.; Marin, G. B.; Broadbelt, L. J. Exploring 1,2-Hydrogen Shift in
52 Silicon Nanoparticles: Reaction Kinetics from Quantum Chemical Calculations and Derivation of Transition
53 State Group Additivity Database. J. Phys. Chem. A 2009, 113, 10933-46.
54 11. Adamczyk, A. J.; Reyniers, M. F.; Marin, G. B.; Broadbelt, L. J. Kinetic Correlations for H(2)
55
56 44 | P a g e
57
58
59
1
2
3 Addition and Elimination Reaction Mechanisms during Silicon Hydride Pyrolysis. Phys. Chem. Chem. Phys.
4 2010, 12, 12676-12696.
5 12. Adamczyk, A. J.; Reyniers, M. F.; Marin, G. B.; Broadbelt, L. J. Hydrogenated Amorphous Silicon
6 Nanostructures: Novel Structure-Reactivity Relationships for Cyclization and Ring Opening in the Gas Phase.
7 Theor. Chem. Acc. 2011, 128, 91-113.
8 13. Kovalenko, M. V.; Manna, L.; Cabot, A.; Hens, Z.; Talapin, D. V.; Kagan, C. R.; Klimov, V. I.;
9 Rogach, A. L.; Reiss, P.; Milliron, D. J. at el. Prospects of Nanoscience with Nanocrystals. ACS Nano 2015, 9,
10
1012-1057.
11
14. Moss, S. J.; Ledwith, A. The Chemistry of the Semiconductor Industry. Chapman and Hall: New York, 1987.
12
15. Petkov, V.; Hessel, C. M.; Ovtchinnikoff, J.; Guillaussier, A.; Korgel, B. A.; Liu, X. F.; Giordano, C.
13
Structure-Properties Correlation in Si Nanoparticles by Total Scattering and Computer Simulations. Chem.
14
15
Mat. 2013, 25, 2365-2371.
16 16. Yoo, S.; Shao, N.; Zeng, X. C. Reexamine Structures and Relative Stability of Medium-Sized Silicon
17 Clusters: Low-Lying Endohedral Fullerene-Like Clusters Si-30-Si-38. Phys. Lett. A 2009, 373, 3757-3760.
18 17. Galashev, A. Y. Molecular Dynamics Study of Hydrogenated Silicon Clusters at High Temperatures.
19 Mol. Phys. 2009, 107, 2555-2568.
20 18. Singh, R. Effect of Hydrogen on Ground State Properties of Silicon Clusters (Sinhm; N=11-15,
21 M=0-4): a Density Functional based Tight Binding Study. J. Phys.: Condens. Matter 2008, 20, 045226.
22 19. Li, C. P.; Li, X. J.; Yang, J. C. Silicon Hydride Clusters Si5H(N) (N=3-12) and their Anions:
23 Structures, Thermochemistry, and Electron Affinities. J. Phys. Chem. A 2006, 110, 12026-12034.
24 20. Zhao, M.; Gimarc, B. M. Strain Energies of Silicon Rings and Clusters. Inorg. Chem. 1996, 35, 5378-
25 5386.
26 21. Raghavachari, K. Theoretical-Study of Small Silicon Clusters - Equilibrium Geometries and
27 Electronic-Structures of Si-2-7, Si-10. J. Chem. Phys. 1986, 84, 5672-5686.
28 22. Bandyopadhyay, D. Study of Pure and Doped Hydrogenated Germanium Cages: A Density
29 Functional Investigation. Nanotechnology 2009, 20, 275202.
30 23. Mahtout, S.; Tariket, Y. Electronic and Magnetic Properties of Crgen (15⩽N⩽29) Clusters: A DFT
31 Study. Chem. Phys. 2016, 472, 270-277.
32 24. Shi, S. P.; Liu, Y. L.; Zhang, C. Y.; Deng, B. L.; Jiang, G. A Computational Investigation of
33 Aluminum-Doped Germanium Clusters by Density Functional Theory Study. Comput. Theor. Chem. 2015,
34 1054, 8-15.
35 25. Kwok, S. H.; Yu, P. Y.; Tung, C. H.; Zhang, Y. H.; Li, M. F.; Peng, C. S.; Zhou, J. M. Confinement
36 and Electron-Phonon Interactions of The E-1 Exciton in Self-Organized Ge Quantum Dots. Phys. Rev. B
37 1999, 59, 4980-4984.
38 26. Headrick, R. L.; Baribeau, J. M.; Lockwood, D. J.; Jackman, T. E.; Bedzyk, M. J. X-Ray and Raman-
39 Scattering Characterization of Ge/Si Buried Layers. Appl. Phys. Lett. 1993, 62, 687-689.
40
27. Schorer, R.; Abstreiter, G.; De Gironcoli, S.; Molinari, E.; Kibbel, H.; Kasper, E. Vibrational
41
Properties of Si/Ge Superlattices - Theory and Inplane Raman-Scattering Experiments. Solid-State Electron.
42
1994, 37, 757-760.
43
28. Weissker, H. C.; Furthmuller, J.; Bechstedt, F. Excitation Energies And Radiative Lifetimes of Ge1-
44
45
xSix Nanocrystals: Alloying Versus Confinement Effects. Phys. Rev. Lett. 2003, 90, 085501.
46 29. Balasubramanian, S.; Ceder, G.; Kolenbrander, K. D. Three-Dimensional Epitaxy: Thermodynamic
47 Stability Range of Coherent Germanium Nanocrystallites in Silicon. J. App. Phys. 1996, 79, 4132-4136.
48 30. Ren, S. F.; Cheng, W.; Yu, P. Y. Microscopic Investigation of Phonon Modes in Sige Alloy
49 Nanocrystals. Phys. Rev. B 2004, 69, 235327.
50 31. Ramos, L. E.; Furthmuller, J.; Bechstedt, F. Quantum Confinement in Si- and Ge-Capped
51 Nanocrystallites. Phys. Rev. B 2005, 72, 045351.
52 32. Viswanathan, R.; Schmude, R. W.; Gingerich, K. A. Molar Atomization Enthalpies and Molar
53 Enthalpies of Formation of GeSi, GeSi2, Ge2Si, and Ge2Si2 by Knudsen-Effusion Mass-Spectrometry. J.
54 Chem. Thermodyn. 1995, 27, 763-770.
55 33. Drowart, J.; Demaria, G.; Boerboom, A. J. H.; Inghram, M. G. Mass Spectrometric Study of Inter-
56 45 | P a g e
57
58
59
1
2
3 Group-Ivb Molecules. J. Chem. Phys. 1959, 30, 308-313.
4 34. Li, S.; Vanzee, R. J.; Weltner, W. Infrared-Spectra of The GeSi and SnSi Molecules in Argon
5 Matrices. Chem. Phys. Lett. 1994, 229, 531-536.
6 35. Andzelm, J.; Russo, N.; Salahub, D. R. Ground and Excited-States of Group Iva Diatomics from
7 Local-Spin-Density Calculations - Model Potentials for Si, Ge, and Sn. J. Chem. Phys. 1987, 87, 6562-6572.
8 36. Sari, L.; Yamaguchi, Y.; Schaefer, H. F. (3)Sigma(-) and (3)Pi States of GeC and GeSi: The
9 Problematic Dissociation Energy of GeC. J. Chem. Phys. 2003, 119, 8266-8275.
10
37. Palummo, M.; Onida, G.; Del Sole, R. Optical Properties of Germanium Nanocrystals. Phys. Status
11
Solidi A 1999, 175, 23-31.
12
38. Jo, C.; Lee, K. Ionization Potentials and Cohesive Energies of Silicon Clusters from the Semi-
13
Empirical Total Energy Tight Binding Method. Phys. Lett. A 1999, 263, 376-381.
14
15
39. Li, S. D.; Zhao, Z. G.; Zhao, X. F.; Wu, H. S.; Jin, Z. H. Structural and Electronic Properties of
16 Semiconductor Binary Microclusters A(m)B(n) (A,B = Si,Ge,C): A B3LYP-DFT Study. Phys. Rev. B 2001, 64,
17 195312.
18 40. Zhu, X. L.; Zeng, X. C. Structures and Stabilities of Small Silicon Clusters: Ab Initio Molecular-
19 Orbital Calculations of Si-7-Si-11. J. Chem. Phys. 2003, 118, 3558-3570.
20 41. Broadbelt, L. J.; Stark, S. M.; Klein, M. T. Computer Generated Reaction Modelling: Decomposition
21 and Encoding Algorithms for Determining Species Uniqueness. Comput. Chem. Eng. 1996, 20, 113-129.
22 42. Wong, H.-W.; Li, X.; Swihart, M. T.; Broadbelt, L. J. Detailed Kinetic Modeling of Silicon
23 Nanoparticle Formation Chemistry via Automated Mechanism Generation. J. Phys. Chem. A 2004, 108,
24 10122-10132.
25 43. Slakman, B. L.; Simka, H.; Reddy, H.; West, R. H. Extending Reaction Mechanism Generator to
26 Silicon Hydride Chemistry. Ind. Eng. Chem. Res. 2016, 55, 12507-12515.
27 44. Curtiss, L. A.; Redfern, P. C.; Rassolov, V.; Kedziora, G.; Pople, J. A. Extension of Gaussian-3
28 Theory to Molecules Containing Third-Row Atoms K, Ca, Ga–Kr. J. Chem. Phys. 2001, 114, 9287-9295.
29 45. Baboul, A. G.; Curtiss, L. A.; Redfern, P. C.; Raghavachari, K. Gaussian-3 Theory using Density
30 Functional Geometries and Zero-Point Energies. J. Chem. Phys. 1999, 110, 7650-7657.
31 46. Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.;
32 Scalmani, G.; Barone, V.; Petersson, G. A.; Nakatsuji, H.; Li, X. at el. Gaussian 16 Rev. B.01, Wallingford, CT,
33 2016.
34 47. Montgomery, J. A.; Frisch, M. J.; Ochterski, J. W.; Petersson, G. A. A Complete Basis Set Model
35 Chemistry. VI. Use of Density Functional Geometries and Frequencies. J. Chem. Phys. 1999, 110, 2822-2827.
36 48. Kalcher, J.; Sax, A. F. Singlet Triplet Splittings and Electron-Affinities of Some Substituted Silylenes.
37 J. Mol. Struct.: THEOCHEM 1992, 85, 287-302.
38 49. Kassaee, M. Z.; Buazar, F. Soleimani-Amiri, S., Triplet Germylenes with Separable Minima at Ab
39 Initio and DFT Levels. J. Mol. Struct.: THEOCHEM 2008, 866, 52-57.
40 50. Apeloig, Y.; Pauncz, R.; Karni, M.; West, R.; Steiner, W.; Chapman, D. Why Is Methylene a Ground
41 State Triplet while Silylene Is a Ground State Singlet? Organometallics 2003, 22, 3250-3256.
42 51. Brulin, Q.; Ning, N.; Vach, H. Hydrogen-Induced Crystallization of Amorphous Silicon Clusters in a
43
Plasma Reactor. J. Non-Cryst. Solids 2006, 352, 1055-1058.
44
52. Scott, A. P.; Radom, L. Harmonic Vibrational Frequencies: An Evaluation of Hartree-Fock, Moller-
45
Plesset, Quadratic Configuration Interaction, Density Functional Theory, and Semiempirical Scale Factors. J.
46
Phys. Chem. 1996, 100, 16502-16513.
47
48
53. Alecu, I. M.; Zheng, J.; Zhao, Y.; Truhlar, D. G. Computational Thermochemistry: Scale Factor
49 Databases and Scale Factors for Vibrational Frequencies Obtained from Electronic Model Chemistries. J.
50 Chem. Theory Comput. 2010, 6, 2872-2887.
51 54. Katzer, G.; Sax, A. F. Beyond the Harmonic Approximation: Impact of Anharmonic Molecular
52 Vibrations on the Thermochemistry of Silicon Hydrides. J. Phys. Chem. A 2002, 106, 7204-7215.
53 55. Vansteenkiste, P.; Speybroeck, V. V.; Verniest, G.; Kimpe, N. D.; Waroquier, M. Applicability of the
54 Hindered Rotor Scheme to the Puckering Mode in Four-membered Rings. J. Phys. Chem. A 2006, 110, 3838-
55 3844.
56 46 | P a g e
57
58
59
1
2
3 56. McQuarrie, D. A.; Simon, J. D. Molecular Thermodynamics. University Science Book, Sausalito, CA:
4 1999.
5 57. Pfaendtner, J.; Yu, X.; Broadbelt, L. J. The 1-D Hindered Rotor Approximation. Theor. Chem. Acc.
6 2007, 118, 881-898.
7 58. Petersson, G. A.; Malick, D. K.; Wilson, W. G.; Ochterski, J. W.; Montgomery, J. A.; Frisch, M. J.
8 Calibration and Comparison of the Gaussian-2, Complete Basis Set, and Density Functional Methods for
9 Computational Thermochemistry. J. Chem. Phys. 1998, 109, 10570-10579.
10
59. Wong, H. W.; Nieto, J. C. A.; Swihart, M. T.; Broadbelt, L. J. Thermochemistry of Silicon-Hydrogen
11
Compounds Generalized from Quantum Chemical Calculations. J. Phys. Chem. A 2004, 108, 874-897.
12
60. Curtiss, L. A.; Raghavachari, K.; Redfern, P. C.; Rassolov, V.; Pople, J. A. Gaussian-3 (G3) Theory
13
for Molecules Containing First and Second-row Atoms. J. Chem. Phys. 1998, 109, 7764-7776.
14
15
61. Geerlings, P.; De Proft, F.; Langenaeker, W. Conceptual Density Functional Theory. Chem. Rev. 2003,
16 103, 1793-1873.
17 62. Pearson, R. G. Absolute Electronegativity and Hardness - Application to Inorganic-Chemistry. Inorg.
18 Chem. 1988, 27, 734-740.
19 63. Parr, R. G.; Pearson, R. G. Absolute Hardness - Companion Parameter to Absolute
20 Electronegativity. J. Am. Chem. Soc. 1983, 105, 7512-7516.
21 64. Parr, R. G.; Donnelly, R. A.; Levy, M.; Palke, W. E. Electronegativity - Density Functional
22 Viewpoint. J. Chem. Phys. 1978, 68, 3801-3807.
23 65. Stein, T.; Autschbach, J.; Govind, N.; Kronik, L.; Baer, R. Curvature and Frontier Orbital Energies in
24 Density Functional Theory. J. Phys. Chem. Lett. 2012, 3, 3740-3744.
25 66. Salzner, U.; Baer, R. Koopmans’ Springs to Life. J. Chem. Phys. 2009, 131, 231101.
26 67. Wielgus, P.; Roszak, S.; Majumdar, D.; Saloni, J.; Leszczynski, J. Theoretical Studies on the Bonding
27 and Thermodynamic Properties of GenSim(m+n) Clusters: The Precursors of Germanium/Silicon
28 Nanomaterials, J. Chem. Phys. 2008, 128, 144305
29 68. Chase, M. W. NIST-JANAF Thermochemical Tables. Fourth Edition, Journal of Physical and Chemical
30 Reference (Monograph 9): 1998.
31 69. Swihart, M.T.; Girshick, S.L. Ab Initio Structures and Energetics of Selected Hydrogenated Silicon
32 Clusters Containing Six to Ten Silicon Atoms. Chem. Phys. Lett., 1999. 307, 527-532.
33 70. Levason, W.; Reid, G.; Zhang, W. Coordination Complexes of Silicon and Germanium Halides with
34 Neutral Ligands. Coord. Chem. Rev. 2011, 255, 1319-1341.
35 71. Gordeychuk, M. V.; Katin, K. P.; Grishakov, K. S.; Maslov, M. M. Silicon Buckyballs versus
36 Prismanes: Influence of Spatial Confinement on the Structural Properties and Optical Spectra of the Si18H12
37 and Si19H12 Clusters. Int. J. Quantum Chem. 2018, 118, 25609.
38 72. Sattler, K.; Nalwa, Hari Singh. Chapter 2 - The Energy Gap of Clusters, Nanoparticles, and Quantum Dots
39 A2 - In Handbook of Thin Films, Academic Press: Burlington, 2002; pp 61-97.
40 73. Oliveira, M. I. A.; Rivelino, R.; Mota, F. D.; Gueorguiev, G. K. Optical Properties and Quasiparticle
41 Band Gaps of Transition-Metal Atoms Encapsulated by Silicon Cages. J. Phys. Chem. C 2014, 118, 5501-5509.
42 74. Mavrovouniotis, M. L. Group Contributions for Estimating Standard Gibbs Energies of Formation
43
of Biochemical-Compounds in Aqueous-Solution. Biotechnol. Bioeng. 1990, 36, 1070-1082.
44
75. Coats, A. M., McKean, D.C., Steele, D. Infrared Intensities of nu3 and nu4 in SiH4, GeH4 and
45
SnH4. J. Mol. Struct. 1994, 320, 269-280.
46
76. Lattanzi, F.; Di Lauro, C.; Horneman, V. M. The ν6, ν8 3ν4+ν12 Infrared System of Si2H6 under
47
48
High Resolution: Rotational and Torsional Analysis. Mol. Phys. 2003, 101, 2895-2906.
49
50
51
52
53
54
55
56 47 | P a g e
57
58
59
1
2
3
4 Table of Contents Graphic
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56 48 | P a g e
57
58
59

Choi 2018

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Choi 2018

Uploaded by

Copyright:

Available Formats

Subscriber access provided by University of Winnipeg Library

A: Spectroscopy, Molecular Structure, and Quantum Chemistry

is published by the American Chemical Society. 1155 Sixteenth Street N.W.,

15 Si1H4 01 34.3 22.57 11.7 31.22 3.1 31.52 2.8

28 Si1H4 01 204.6 204.38 0.2 204.56 0.0 204.48 0.1

34 Ge1H4 01 217.1 217.31 -0.2 217.48 -0.4 217.41 -0.3

50 Ge3H8 01 N.A. 126.17 - 128.35 - 126.92 -

Heat of Formation with BAC (KJ/mol)

Heat of Formation with BAC (KJ/mol)

Heat of Formation with BAC (KJ/mol)

You might also like